The Management of Productivity and Technology in Manufacturing PDF

The Management of
Productivity and
Technology in
Manufacturing
The Management of
Productivity and
Technology in
Manufacturing
Edited by
Paul R. Kleindorfer
Center for the Study of Organizational Innovation
The Wharton School
University of Pennsylvania
Philadelphia, Pennsylvania
Plenum Press • New York and London

Library of Congress Cataloging in Publication Data
Main entry under title:
The Management of productivity and technology in manufacturing.
Includes bibliographical references and index.

1. Production management-Congresses. 2. Robots, Industrial-Congresses. 3. Flexible
manufacturing systems-Congresses. 4. Manufacturing processes- Technological innovations
-Congresses.!. Kleindorfer, Paul R.
TS155.A1M356 1985 658.5'14 85-17022
ISBN-13: 978-1-4612-9516-7 e-ISBN-13: 978-1-4613-2507-9

DOl: 10.1007/978-1-4613-2507-9
Commissioned papers presented at the Wharton Conference on Productivity,

Technology, and Organizational Innovation, held December 8-9, 1983, in
Philadelphia, Pennsylvania, under the sponsorship of the Center for the Study of
Organizational Innovation, University of Pennsylvania
©1985 Plenum Press, New York

Softcover reprint of the hardcover 1st edition 1985
A Division of Plenum Publishing Corporation

233 Spring Street, New York, N.Y. 10013
All righ ts reserved
No part of this book may be reproduced, stored in a retrieval system, or transmitted

in any form or by any means, electronic, mechanical, photocopying, microfilming,
recording, or otherwise, without written permission from the Publisher
CONTRIBUTORS
Thomas Allen • Sloan School of Management, Massachusetts Institute of

Technology, Cambridge, Massachusetts
Rajiv D. Banker. School of Urban and Public Affairs, Carnegie-Mellon
University, Pittsburgh, Pennsylvania
Jay B. Barney. Graduate School of Management, University of California,

Los Angeles, California
Morris A. Cohen • The Wharton School, University of Pennsylvania,

Philadelphia, Pennsylvania
Christopher Fuselier. General Electric Company, Industrial Automation

Systems Department, Albany, New York
Ralph Katz • Graduate School of Business, Northeastern University,

Boston, Massachusetts
Linsu Kim. Korea University, Seoul, Republic of Korea
Hau L. Lee. Department of Industrial Engineering and Engineering

Management, Stanford University, Stanford, California
Steven Miller • Graduate School of Industrial Administration, Carnegie-

Mellon University, Pittsburgh, Pennsylvania
Jack Romanosky. Arthur Andersen & Company, Philadelphia,

Pennsylvania
Charles S. Skinner • Strategic Technology Inc., Cleveland, Ohio
John F. Tomer • Department of Economics and Finance, Manhattan

College, Riverdale, New York
James M. Utterback. Massachusetts Institute of Technology, Cambridge,

Massachusetts
Steven A. Weiner. Ford Motor Company, Detroit, Michigan
v
PREFACE
This volume is concerned with the nature of new manufacturing

technologies, such as CAD/CAM and robotics, as well as ap-
propriate methodologies for evaluating whether such technologies are
financially and organizationally viable in particular contexts. The
chapters included here were commissioned as papers for presen-
tation at The Wharton Conference on Productivity, Technology, and
Organizational Innovation, which took place in Philadelphia on
December 8 and 9 of 1983. The conference was sponsored by The
University of Pennsylvania's Center for the Study of Organizational
Innovation.
There has been a surge of interest in the area of manufacturing
over the past ten years as managers have come to realize that the
operations function is critical to remaining competitive. New status
has been given to factory and operations managers. New programs
revitalizing manufacturing and distribution have been introduced in
organizations. Corporate strategy is now explicitly considering
operations and manufacturing functions. And the curricula of leading
business schools are reflecting the rapidly advancing research on
technology management and manufacturing operations.
In spite of these important signs of progress, we are clearly just
at the beginning of understanding the issues involved here. The
present volume provides a state-of-the-art review of the realities of
technology management and manufacturing strategy. As described in
the Editor's Introduction, we address four topics: The Nature of New
Manufacturing Technology, Innovation and Manufacturing Strategy,
Productivity Management, and Technology Management and Organ-
ization. These issues are clearly very important themes for U.S.
industry and its policymakers, given the tremendous cost of new
information-intensive technologies and the implications of these for
any firm's work force and support staff.
I am pleased to acknowledge the contributions of industry
representatives of The Wharton Partnership. I am also delighted to
Vtl
V," PREFACE
acknowledge the financial support of the Continental Group, Texaco

Corporation, and the Gulf Oil Foundation. Finally, a special note of
thanks is due to the participants of the conference, whose construc-
tive comments on early versions of the chapters presented herein
contributed significantly to a lively conference and a much improved
final product.
PAUL R. KLElNOORFER
Director, Center for the Study
of Organizational Innovation
Philadelphia
CONTENTS
Editor's Introduction 1
1. The Nature of New Manufacturing Technology................... 1
2. Innovation and Manufacturing Strategy............................... 2
3. Productivity Management.................................................... 3
4. Technology Management and Organization ......................... 5
References................................................................................ 6
I. The Nature of New Manufacturing Technology
CHAPTER 1
Industrial Robotics and Flexible Manufacturing Systems:
An Overview 9
Steven M. Miller
1. Introduction ........................................................................ 9
2. Background on Robotic Manipulators.................................. 10
2.1. The Robotic Manipulator ................................ ... ... ... .... 10
2.2. Defining Levels of Robot Technology Based on
Internal vs. External Sensing......................................... 11
2.3. Limitations of Level I Robots ...................................... 14
3. Extensions in the Capabilities of Insensate Robots ............... 21
3.1. Increased Availability of High-Precision Robots for
Light Manufacturing.................................................... 21
3.2. Increased Availability of Computer-Controlled Robots 23
3.3. The Integration of Sensory Information Processing
with Robotic Devices................................................... 25
3.4. Commercial use of Machine Vision Systems Since 1979 31
4. Flexible Manufacturing Systems........................................... 33
4.1. An Overview of Manufacturing Systems ...................... 33
4.2. Flexible Manufacturing Systems for Machining ............ 39
IX
x CONTENTS
4.3. Developments in Flexible Manufacturing Systems ........ 46

Notes....................................................................................... 50
References................................................................................ 52
CHAPTER 2
Perspectives on Automotive Manufacturing 57
Steven A. Weiner
1. Introduction ........................................................................ 57
2. Scope of Ford North American Manufacturing Operations 58
3. Vehicle Assembly Operations .............................................. 59
4. Stamping Operations ........................................................... 62
5. Scheduling ...................... ........... .......................................... 64
6. "New" Operating Strategies in Automotive Manufacturing 67
7. Summary ............................................................................. 70
References................................................................................ 71
CHAPTER 3
Impacts of Robotic and Flexible Manufacturing Tech-
nologies on Manufacturing Costs and Employment 73
Steven M. Miller
1. Introduction........................................................................ 73
2. Metalworking Industries ...................................................... 74
3. The Impacts of Robotic Manipulators .................................. 76
3.1. Calculation of Payback Periods Based on Direct Labor
Savings......................................................................... 82
3.2. The Price Elasticity Argument ..................................... 90
4. The Impacts of the Fully Utilized, Flexibly Automated
Factory for Batch Production ...... ................ ........................ 94
Notes....................................................................................... 107
References................................................................................ 108
II. Innovation and Manufacturing Strategy
CHAPTER 4
Invasion of a Stable Business by Radical Innovation 113
James M. Utterback and Linsu Kim
1. Introduction ...... .................................................................. 113
CONTENTS Xl
2. A model of Innovative Challenges and Defensive Responses 121

2.1. The New Technology Enters a Market Demanding
New Performance ........................................................ 121
2.2. A General Pattern of Response to Radical Innovation .. 123
3. Application of the Model to Specific Examples..................... 133
3.1. Product-Process Discontinuities .................................. 133
3.2. Product Discontinuities................................................ 137
3.3. Process Discontinuities ............................. ......... ... ....... 141
3.4. Process-Product Discontinuities .................................. 143
4. Some Concluding Observations ........................................... 146
Notes....................................................................................... 149
References................................................................................ 149
CHAPTER 5
Manufacturing Strategy: Concepts and Methods 153
Morris A. Cohen and Hau L. Lee
1. Introduction ........................................................................ 153
2. Manufacturing Strategy: Decisions and Performance ........... 155
2.1. Product Decisions........................................................ 156
2.2. Process Decisions......................................................... 159
2.3. Facility Decisions......................................................... 164
2.4. Control Decisions........................................................ 165
2.5. Organization Decisions................................................ 167
3. The Manufacturing Strategy Paradigm ................................. 168
4. MPSS Functions and Design ................................................ 172
4.1. Segmentation Analysis ................................................. 172
4.2. Cost/Performance Analysis ......................................... 174
4.3. Production/Distribution Model................................... 175
4.4. Production Process Model........................................... 179
4.5. Market Response ModeL............................................ 182
5. Implementation Considerations ..... '" ...................... '" ... ....... 183
References................................................................................ 186
CHAPTER 6
The Executive Dilemma: New Industrial Automation Systems 189
Christopher S. Fuselier
xii CONTENTS
III. Productivity Management
CHAPTER 7
Working Smarter the Japanese Way: The X-Efficiency of
Theory Z Management 199
John F. Tomer
1. Introduction ........................................................................ 199
1.1. X-Efficiency and Productivity ...................................... 202
1.2. Plan of the Chapter ...................................................... 203
2. Implicit Psychological Contracts, Transaction Costs, and
Cooperation.............. ... .................... ....... ... ...... ................... 204
2.1. The Implicit Psychological Contract ............................ 204
2.2. The IPC Hypothesis ........................... ......................... 206
3. The Determinants of Effort ................................................. 210
3.1. The Utility-Effort Relation.......................................... 210
4. Conclusion .......................................................................... 218
Notes....................................................................................... 221
References................................................................................ 224
CHAPTER 8
Theory Z, Institutional Economics, and the Theory of
Strategy 229
Jay B. Barney
1. Introduction........................................................................ 229
2. Microeconomics and Type Z Management ........................... 229
3. Institutional Economics, Management, and the Theory of
Strategy............................................................................... 233
Notes....................................................................................... 236
References................................................................................ 236
CHAPTER 9
Productivity Measurement and Management Control 239
Rajiv D. Banker
1. Introduction.............................. ........................................... 239
2. Congruence with Profit Maximization ................................. 242
3. A Framework for Productivity Analysis .............................. 245
CONTENTS xm
4. Concluding Remarks ........................................................... 253

References................................................................................ 254
CHAPTER 10
The 1983 Global Manufacturing Futures Study Preliminary
Report Synopsis 259
Jack Romanosky
1. Introduction.......... ..... ......................................................... 259
2. Conclusions of the Survey.......................... .................. ....... 260
3. Major Observations ............................................................. 260
3.1. Strategies and Competitive Priorities ............................ 261
3.2. Major Concerns ........................................................... 263
3.3. Current Practices ......................................................... 265
3.4. Programs for the Future............................................... 266
4. Environmental Observations................................................ 270
IV. Technology Management and Organization
CHAPTER 11
Organizational Issues in the Introduction of New
Technologies 275
Ralph Katz and Thomas J. Allen
1. Introduction........................................................................ 275
2. Problem Solving, Communications, and the Mobility of
People ................................................................................. 278
2.1. Problem-Solving Processes............ .......... ...... ............... 280
2.2. Communication and Information Processing................ 282
2.3. Cognitive Processes ..................................................... 283
2.4. Mobility of People and the "Not Invented Here"
Syndrome .................................................... ................ 285
3. Organizational Structures .................................................... 287
4. Organizational Controls ...................................................... 292
4.1. Technical Understanding.............................................. 292
4.2. Technical Feasibility .................................................... 294
4.3. Research and Development Overlap............................. 294
4.4. Growth Potential......................................................... 295
xiv CONTENTS
4.5. Organizational Slack and Sponsorship.......................... 296

4.6. Organizational Rewards............................................... 298
References................................................................................ 299
CHAPTER 12
The Strategic Management of Technology 301
CharLes S. Skinner
1.Technology in the 1970s ...................................................... 301
2.Impact of Technology on Business in the 1980s ................... 306
3.Booz-Allen's View of Technology Management .................. 309
4.Booz-Allen's Approach to Technology Strategy.................. 311
4.1. Developing the Technology Portfolio........................... 311
4.2. Matching Technology and Business Portfolios.............. 314
4.3. Setting Technology Investment Priorities ..................... 315
Notes....................................................................................... 316
Subject Index 317
Author Index 329

EDITOR'S INTRODUCTION
The chapters in this volume are concerned with four related

areas:
(1) the nature of new manufacturing technology;
(2) innovation and manufacturing strategy;
(3) productivity management; and
(4) technology management and organization.
Little need be said to demonstrate the importance of these issues

for practicing manufacturing managers. Put simply, information
technology and the creative destruction of international competition
have jointly revolutionized manufacturing in the past two decades.
New methods for material handling and integration of numerically
controlled machines led the way in the 1960s. By the mid-1980s,
every facet of manufacturing operations and distribution has been
affected by new computer-aided planning, design, and control
procedures. To remain competitive in today's as well as tomorrow's
economic environment, the manufacturing manager must understand
the nature of these new information-intensive technologies and when
and where it is profitable to implement them. This volume assembles
recent research findings of U.S. scholars and practitioners on these
issues. Related research on European and Japanese experience is
available in Yoshikawa, Rathmill, and Hatvany (1981) and Vlietstra
and Wielinga (1983).
This book is organized into four parts corresponding to the four
areas of concern. A brief overview of each of these sections follows.
1. THE NATURE OF NEW MANUFACTURING

TECHNOLOGY
Part I deals with the nature of the new manufacturing technolo-

gies which have been the focus of discussion. These include, of
course, a spectrum of mechanical, informational, and organizational
2 INTRODUCTION
innovations. The first chapter by Steven Miller discusses the nature of

industrial robotics and their integration in flexible manufacturing
systems. Miller's paper begins with a survey of available technology
and current applications of robotic manipulators and provides an
overview of manufacturing systems. On the basis of this framework,
Miller considers various approaches to integrating robotics technol-
ogy into controllable, flexible manufacturing systems, including the
use of artificial intelligence and sensor-based systems. A sample of
current FMS applications in the U.S. and Japan, with corresponding
financial and performance statistics, completes this paper and prov-
ides a nice introduction to the forces shaping the current revolution
in manufacturing technology.
In the second chapter, Steven Weiner describes the use of new
manufacturing technologies at the Ford Motor Company. Weiner's
chapter provides an overview of automotive manufacturing, an
industry which will doubtless continue to be a bellwether and testing
ground for manufacturing technology. Using Ford as a specific case,
Weiner considers the implementation requirements for such new
operating strategies as computer-integrated manufacturing, just-in-
time inventory control, and statistical process control.
The final chapter in Part I is again authored by Steven Miller and
concerns the likely impacts of new manufacturing technologies
(NMT) on unit costs and employment. Concentrating on the
metal-working sector of U.S. manufacturing industry, Miller reviews
the growing literature on cost/ employment effects of NMT. This
review and a coupled cross-sectional statistical analysis provide
insights on interactions between cost, utilization, and output. Miller
argues on the basis of his analysis that NMTs will directly affect a
large proportion of the work force in manufacturing and that
significant per-unit cost savings will materialize from these
technologies.
2. INNOVATION AND MANUFACTURING STRATEGY
The traditional evolution of product and process interactions is

from customized job-shop to high-volume/assembly line. The
INTRODUCTION 3
chapter by Utterback and Kim extends this discussion to include the

effects of a radical innovation, such as CAD/CAM, on the traditional
evolution. The authors develop hypotheses for the reactions of both
the innovating firm and defending firms who continue to use the
existing technology. The resulting strategic scenario, and its com-
petitive evolution, are illustrated with many historical cases.
The issue of competitive strategy is pursued further in the
chapter by Cohen and Lee. They take on the demanding task of
delineating an operational framework for manufacturing strategy in
support of competitive strategy. Their analysis provides a conceptual
framework for manufacturing strategy which incorporates technol-
ogy and factory-focus with managerial decisions and strategic perfor-
mance criteria such as cost, quality, flexibility, and dependability.
They also discuss the formulation of a- manufacturing planning
support system to implement their framework.
The final chapter in Part II is by Christopher Fuselier of the
General Electric Company, a company which has made large
investments in industrial automation systems. Fuselier presents
several case studies from General Electric's own operations to
indicate two important characteristics of flexible manufacturing
systems. First, they are quite expensive (from a few tens of millions
to several hundreds of millions of dollars), and, second, they can
provide immense benefits in cost, quality, and flexibility. The key
problem addressed by Fuselier is how to justify the investment cost
required for such systems by the uncertain benefits they may
provide.
3. PRODUCTIVITY MANAGEMENT
Following Ben Franklin's adage, "There is nothing so galling as

a good example," the Japanese experience in introducing new
management techniques and technologies in manufacturing has been
both a source of inspiration and frustration to others. Several
interesting research questions have emerged from this process,
including the nature of Japanese management techniques and the
extent of their transferability elsewhere. These are taken up in the
4 INTRODUCTION
chapter by John Tomer, who concentrates on the incentives for

performance and self-monitoring which various management struc-
tures imply. Tomer argues that the major reason for the success of
Japanese firms in the manufacturing area has been the increased
productivity resulting from the long-term performance incentives
which their style of organization and management entail.
The chapter by Jay Barney was presented as a "discussant's
comment" on the Tomer chapter, but it is reproduced here because it
raises several additional points of interest concerning the strategic
ramifications of management. Barney's main point is that the benefits
arising from Japanese style management may depend strongly not
only on cultural conditions but also on market structure. He
emphasizes that management strategy, like other elements of com-
petitive strategy, should be directed towards exploiting comparative
advantages of the firm. He goes on to explore the implications of this
for Japanese style management as a strategic option involving, inter
alia, significant investments in the work force. Barney argues that this
would be a sensible strategy only if returns on this investment are
attractive, and he points out several determinants of such returns.
The chapter by Rajiv Banker addresses the accounting and
management control issues associated with productivity. Banker
points out that effective productivity control requires some theoreti-
cal, historical or comparative benchmark. These benchmarks repre-
sent either the status quo or some desired point in measuring the
effectiveness of input to output transformation (i.e., of productivity).
After a general introduction on the current state of accounting theory
for productivity measurement, Banker describes a new framework
for management control of productivity and illustrates its use in areas
like traditional variance analysis.
A further contribution to the productivity management debate is
the chapter by Jack Romanosky, who provides a practitioner's
perspective on the issues involved in incorporating productivity in
manufacturing strategy. Romanosky summarizes the findings of a
survey of over 600 business units from North America, Europe, and
Japan on similarities and differences in manufacturing priorities and
plans for achieving these. Romanosky points out that there is a
significant agreement across these countries on priorities for the
INTRODUCTION 5
future, but there exist significant differences on the best means of

achieving these priorities, with relative differences in emphasis on
such issues as investments in process technology, software, and
quality control. These international comparisons confirm the intricate
relationship between organizational and cultural context and manu-
facturing priorities and strategies. They also underline continuing
interest and importance of international comparative research on
manufacturing organizations.
4. TECHNOLOGY MANAGEMENT AND ORGANIZATION

The final two chapters in this volume address the specific
question of technology management. Ralph Katz and Thomas Allen
first consider in their chapter the organizational issues associated
with the introduction of new technologies. It has long been
recognized that organization design and behavior are the result of
complex interactions of organizational participants with authority
structure, reward systems, information systems, and technology.
Given the importance of new technologies in the current manufac-
turing context, however, it is of some interest to focus specifically on
organizational consequences of technology changes. Katz and Allen
focus especially on the difficult problem of designing organizations
which simultaneously function effectively in an operational sense on
today's products and processes, while maintaining a sense of dynam-
ism and innovation which will allow their continuing growth and
viability in the future. Katz and Allen provide some solace and
wisdom for those beleaguered managers attempting to balance these
conflicting objectives in organizations, such as those in the manufac-
turing sector, currently undergoing radical technological change.
The final chapter in this volume is by Charles S. Skinner former
Vice President of Booz-Allen & Hamilton. Skinner provides an
overview of Booz-Allen's technology management approach, con-
centrating on the integration of this with competitive and manufac-
turing strategy. Skinner provides a basic framework for linking the
corporate business portfolio with its technology portfolio and
organizational and market opportunities. The discussion is illustrated
by several recent case histories in manufacturing.
6 INTRODUCTION
The intent of the conference where the above chapters were

presented as papers was to cover the broad spectrum of current issues
facing u.s. manufacturing industries, as they face sweeping changes
in technology and management the likes of which have not been
witnessed since the industrial revolution. This broad spectrum
certainly includes the basic issues of technology, productivity, and
organizational innovation, which were the organizing themes of the
conference itself. As the chapters in this volume indicate, however,
these issues spawn many other concerns which scholars, managers,
and public policy makers are just beginning to understand. In this
sense, the papers presented here are only a beginning. Much
theoretical and empirical research remains to be done. What is
encouraging and evident in this volume, however, is the level of
interaction and cross-fertilization between theory and practice which
these issues evoke.
REFERENCES
Vlietstra, J., and Wielinga, R. F. (eds.), Computer-Aided Design. North-

Holland, Amsterdam, 1983.
Yoshikawa, H., Rathmill, K., and Hatvany, J., Computer-Aided
Manufacturing: An International Comparison. National Academy Press,
Washington, D.C., 1981.
PART I
THE NATURE OF
NEW MANUFACTURING
TECHNOLOGY
CHAPTER 1
INDUSTRIAL ROBOTICS AND

FLEXIBLE MANUFACTURING
SYSTEMS
An Overview
Steven M. Miller
1. INTRODUCTION
This chapter reviews some of the recent developments in

robotics and flexible manufacturing technology. Attention if focused
on two broad classes of developments:
1. Extensions of the capabilities of commercially available robo-

tic manipulators;
2. Integration of robots, machine tools, sensory information
processing system, parts-handling and transport devices, and
computers into manufacturing systems.
Topics covered that relate to extensions of commercially avail-

able robotic manipulators include (1) increases in the number of
robot models designed for precision placement, (2) increases in the
use of computer-controlled robots, and (3) the integration of sensory
information processing with robotic devices. Topics covered that
relate to the integration of machines into flexible manufacturing
systems (FMS) include (1) the integration of multiple FMS for
Steven M. Miller. Graduate School of Industrial Administration, Carnegie-Mellon

University, Pittsburgh, Pennsylvania 15213.
9
10 STEVEN M. MILLER
machining in one plant, (2) the use of FMS in manufacturing

processes other than machining, and (3) the addition of artificial
intelligence and sensor-based systems to FMS.
The chapter begins with a definition of robots and a discussion of
limitations of commercially available robot manipulators. These
preliminaries provide a foundation for appreciating recent
developments.
2. BACKGROUND ON ROBOTIC MANIPULATORS
2.1. The Robotic Manipulator

An industrial robot is essentially a device that can move
materials, parts, or tools from one point in space to another under
programmed control without human intervention. Tracing the cur-
rent industrial robot's roots back to both the teleoperator and the
numerically controlled (NC) machine tool, Paul (1981) refers to a
robot as "a device combining the articulated linkage of the teleopera-
tor with the servo axis of the numerically controlled milling
machine." A robot is distinguishable from a teleoperated mechanical
arm in that it is programmable and can execute its program
automatically. By contrast, the teleoperator serves as an extension of
the human controller, who is either attached to it physically, or
remotely via radio waves.
The robot is able to manipulate parts with an "end effector"-a
mechanical gripper-that can grasp an object, pick it up, reorient it,
and reposition it. The end effector is attached to a mechanical arm
that can be instructed to move from one position to the next to pick
up and move things. Thinking of a robot as a programmable
manipulator seems to easily distinguish a robot from a numerically
controlled machine tool, but the difference becomes ambiguous on
closer inspection. Both the robot and the NC tool have several
independent axes of motion. The types of servo mechanisms used to
control the motion along each axis are similar in each case. Both can
be programmed to move through a specified cycle. Once set up and
programmed, both the robot and the NC tool can complete the
INDUSTRIAL ROBOTICS 11
designated task without human intervention. While a robot is

generally recognized for its ability to pick up an object and
reposition it, some NC tools can effectively perform these functions,
although not so easily. While an NC tool does not have a gripper,
some can still acquire parts from a specially designed pallet and move
the pallet onto the machine bed. The machine bed may also have
several degrees of freedom, allowing the part to be moved and
rotated. A robot must also be multifunctional in that it can
manipulate an object in different ways and manipulate different types
of objects. Yet a sophisticated machine tool can do the same, since it
can also be programmed to move its workhead to carry various types
of tools. An NC machine tool is used to alter the physical state of an
object by changing its shape or adding or subtracting to its material
content. By controlling the position of a tool, a robot can also be
programmed to alter the physical state of an object. These com-
parisons show that a robot basically is a type of numerically
controlled machine tool. However, a robot is usually more versatile
in grasping, manipulating, and repositioning objects than a conven-
tional (e.g., metal-shaping) machine tool because it is designed
principally to alter the spatial state of an object.
2.2. Defining Levels of Robot Technology Based on Internal vs.

External Sensing
A useful way to delineate levels of robot sophistication is to
view a robot as an information-processing system. The sophistication
of the system depends on two factors: (1) where it acquires
information from and (2) how it processes the information it
acquires. Any robot can acquire some types of information from its
own internal mechanisms, including its mechanical, electrical, or
electronic subsystems. Examples are whether the robot is "on" or
"off", the temperature of its various subcomponents, and whether its
gripper is open or closed. Servo-controlled robots also monitor
variables such as the position of a particular joint in the arm or wrist
and the load on a particular joint. Our first concern is to distinguish
between a robot that can only obtain information from its own
internal workings and a robot that can also obtain information from
12 STEVEN M. MILLER
its external environment. We refer to the former as an insensate robot

and refer to the latter as a sensate robot.
Various types of information can be acquired from the external
environment. The simplest example is a binary signal (e.g., absent or
present, yes or no, on or off, positive or negative). For example,
when a part advances to the end of a conveyor, it may push against a
switch which sends a "present" signal to the robot. The signal is used
to sequence prespecified programs. We use the term "Level I" to
refer to the insensate robot defined above, as well as to the robot that
can sense externally generated binary signals and use them to
sequence predetermined programs. A definition is given below:
Level I Robot. A robot system where the goals cannot be
modified through the result of a feedback and control process that
senses the external environment. All aspects of the task must be
explicitly specified in advance. Predetermined alternatives may be
invoked via limit switches, timing devices, or simple transducers
which transmit a binary inputs or measure predetermined thresholds.
An example of complex sensing and processing of external
information would be the use of visual information to identify a part
on a conveyor, to determine its location, and finally, to have the
robot move to this location and grip the part. We refer to such a
robot as a Level II robot. It can sense more than a single binary
signal and make decisions more complex than sequencing prespeci-
fied alternatives. A definition is provided below:
Level II Robot. A robot system where the goals can be
modified through the result of an adaptive feedback and control
process that senses the external environment via some type of
transducer. Instruction can be adapted and generated in response to
changes in the external environment. Thus, it is possible to perform
some kinds of tasks without having to explicitly specify all of its
aspects in advance.
Capabilities generally required within each of the major catego-
ries of robot applications in manufacturing are shown in Table 1.
Three types of capability are distinguished: (1) the ability to
transport an object from one place to another, (2) the ability to
manipulate an object, and (3) the ability to use external sensing for
feedback control. The table indicates that a robot that can only
TABLE I
Major Categories of Robot Applications in Manufacturing and Capabilities
Generally Required Within Each Application a
Major Examples Capabilities required to perform application

application within
area area Transport Manipulation Sensing
Material Parts handling x

handling Palletizing x
Transporting x
Heat treating X
Machine Die cast machines X X

loading Automatic presses X X
NC milling machines X X
Lathes X X
Spraying Spray painting X
Resin application X
Welding Spot welding X

Arc welding X
Machining Drilling X X
Deburring X X
Grinding X X
Routing X X
Cutting X X
Forming X X
Assembly Mating parts X x

Acquiring parts X X
Inspection Position control X

Tolerance X
a Source: Tech Tran Corp. (1983, p. 63).
transport and manipulate objects (Level I robot) is usually adequate

to perform material handling, machine loading, spraying, and weld-
ing tasks. A robot with external sensing, however, could also perform
machining (tool positioning), assembly, and inspection tasks. This is
only a rough guide, since there are clearly applications in material
handling, machine loading, spraying, and welding that require (or
could benefit from) external sensing and applications in machining,
assembly, and inspection that can be engineered to eliminate the need
14 STEVEN M. MILLER
for external sensing, Nonetheless, the table conveys the general

notion that the use of external sensing would broaden the range of
manufacturing tasks that could be performed by a robot.
2.3. Limitations of Level I Robots

To understand the limitations of Level I robots, we focus on the
elementary task of moving the robot's end effector from one position
to another. It is surprisingly difficult for an insensate robot to
perform this basic function. To begin with, instructing the robot
where to move its gripper is a cumbersome process. With "walk-
through" programming, an operator takes hold of the end effector
and moves it through a specific continuous path while recording the
pattern of motion. With "lead-through" programming, an operator
commands the robot to move, using a teach pendant or a joy stick
(similar to those used in video games), recording a series of discrete
points. Both programming procedures require that the robot be taken
out of production each time it is taught a new task. Once the
program is recorded, however, it can be be quickly recalled.
Transmitting new instructions to a robot by inputing coordinates
into a computer (even while the robot is in operation) is becoming
more widespread and is discussed in the next chapter. Another
problem is that after an insensate robot carries out its instructions, it
does not know the precise position of its most critical component, its
end effector. In short, it is cumbersome to tell a conventional robot
where to go, and after it moves, it does not know if it went exactly to
the right place.
A schematic of the position control system for an insensate
robot (with servo control), shown in Figure 1, is useful in illustrating
limitations. The position of the tip of the end effector in the real
world is given by x" the reference position. While programming the
robot, the operator thinks in terms of the position of the end effector
in (real) world coordinates. The robot system is actually recording
the position and angles of each of the separate joints within the arm
in its own internal coordinate system. The robot controller contains a
model consisting of a set of linked equations that specifies the
position of the end effector as a function of the rotations and
Load Disturbance
Robot Model
x;
••
• End-Effector
Kinematic Position
Arm Solution
Reference
Position
Joint Control
FIGURE 1
Positioning system for an insensate robot with servo control. Conventional (insensate)
manipulator position control system. An arm solution maps the reference end-effector
position into coordinates for each of the individual joints. Closed-loop joint
controllers control the positions of each joint independently. There is no direct
closed-loop feedback between the final position of the end effector and the reference
position. (From Sanderson and Weiss, 1983, p. 109.)
pOSltIOnS of all the joints. The joint motions required to achieve a

desired end effector position can be calculated by solving these
equations. This process is represented by the box labeled "kinematic
arm solution." Once all the reference positions are specified by the
operator, the robot must repeat the desired sequence of motions. The
box labeled "closed-loop joint control" represents this process. Each
of the separate joints is moved to the positions given by the robot
model and kinematic arm solution. Using potentiometers, optical
encoders, and similar devices, the current position of the joint is
sensed and compared to the desired position. Each joint is moved
until the difference between the current position and the desired
position is reduced to zero.
There are load disturbances at each joint, as indicated in the
figure. Internal parts, such as gears or linkages, can wear and
introduce inaccuracies into the positioning process. Also, physical
forces, such as friction and backlash, cause distortions that must be
compensated for in the calculations. The effects of these disturbing
forces can only be approximated and cannot be completely elimi-
nated. The final end effector position, given by X in the figure, is not
independently measured and is therefore not known by the robot
16 STEVEN M. MILLER
itself. It is assumed that by approximately positioning the individual

joints in the manipulator arm, the end effector will end up in the
desired position. Because of inaccuracies in the robot model (the
original specification of the joint movements required to achieve the
programmed reference position) and because of many types of
disturbances, the end effector is not always exactly in the predicted
position. This is known as open-loop control. It is characteristic of an
insensate robot. More detailed discussions of the basic control and
structure of the insensate industrial robot are given in Paul (1981),
Luh (1983), and Coiffet (1983).
Even with their inherent limitations, almost all of the robots in
the field today are insensate. In his description of the earliest
machines manufactured by Unimation in the 1960s, Paul (1981, p. 2)
explains how such a robot is typically utilized:
The industrial robot could be taught to perform any simple

job by driving it by hand through the sequence of task positions,
which were recorded in digital memory. Task execution con-
sisted in replaying these positions by servoing the individual
joint axes of the robot. Task interaction was limited to opening
and closing of the tongs or end effector, and to signaling external
equipment or waiting for a synchronizing signal. The industrial
robot was ideal for pick and place jobs such as unloading a
diecasting machine. The part would appear in a precise position,
defined with respect to the robot; it would be grasped, moved
out of the die, and dropped on a conveyor. The success of the
industrial robot, like the numerically controlled (NC) milling
machine, relied on precise, repeatable digital servo loops. There
was no interaction between the robot and its work. If the diecast
machine were moved, the robot could in no way adapt to the
new position, any more than an NC milling machine could
successfully cut a part if the stock were arbitrarily relocated
during cutting. If the diecasting machine were moved, the robot
could, however, be retaught. The success of the industrial robot
lay in its application to jobs in which task positions were
absolutely defined, and in its reliability and positioning repeata-
bility in lieu of adaptation.
In order that task pOSitIOnS be absolutely defined with no

unplanned variations, the workplace within reach of the robot had to
be carefully controlled. The requirement for workplace structuring to
use insensate robots was emphasized in an assessment of robot
technology report prepared by the Eikonix Corporation in 1979
(Eikonix Corp, 1979):
Present day industrial robots have achieved their most

successful application in such areas as machine loading and
unloading, particularly die casting machines, and spot welding in
the automobile industry. All of these applications require
complete structuring of the environment in order that the
appropriate reference of motions and functions can be specified
before-hand. While these robots can accommodate changes,
these changes have to be completely specified in all their
detail ...
Planning for variations requires a complete structuring of
the environment in which the robot is to function. The
environment need not be totally static, but all aspects of its
dynamics must be explicitly known. This structuring of the
environment significantly adds to the cost of using a robot. Not
only must physical modifications be made, but the production
engineers must come to an explicit understanding of the
relationship between task and the environment in ways that are
totally unnecessary when human workers are used. 1
Much engineering effort and additional accessory hardware is

required to ensure that all aspects of the task can be prespecified in
complete detail. For example, a robot programmed to move to a
specific location and pick up a part requires parts feeding and
orientation devices to ensure that each part arrives at the same
position in the same orientation. A part slightly out of position or
orientation might bring the system to a stop. Grippers and motion
sequences for gripping parts have to be designed to compensate for
uncertainties in the precise position of the end effector. When a robot
is required to position a tool precisely with respect to a workpiece (as
in arc welding), it is often necessary to secure the workpiece to a
table that can be precisely positioned, minimizing the requirements
18 STEVEN M. MILLER
for moving the robot arm. The tools themselves often have to be
specially designed so that they can be held by the robot.
As a result of these requirements, the cost of an insensate robot
is typically only a fraction of the total cost of an application. A recent
estimate of the cost of a typical robot application is shown in Figure
2 for six different types of applications. The basic robot cost includes
the cost of the arm, the power supply, and the controller. According
to this estimate, the basic robot typically comprises only 40% of the
cost in an assembly application, 55% of the cost in a machine loading
or welding application, and 70% of the cost in a spray painting
application. The "accessories cost" includes optional equipment,
special tools, grippers, and maintenance and test equipment. Installa-
tion includes site preparation, work rearrangement, utility connec-
tions, and interfaces. These other types of expenses account for the
remainder of the total installation cost. The estimates do not include
costs required for the study and selection of the application to be
robotized (planning) or for selecting the robot to be used (applica-
tions engineering). Planning and applications engineering work
required before a robot is purchased can add tens of thousands of
dollars to the total cost.
Other studies suggest that the basic robot accounts for even a
smaller percentage of total cost than suggested by Figure 2. In one
study of robot use, which focused on applications in the auto
industry, Hunt and Hunt (1983) reported that the basic (insensate)
robot usually represented less than 40% of the total costs in
applications areas such as welding, where many robots have already
been installed. Miller (1983) estimated that in machine loading and
unloading, the basic robot typically comprises only 20%-30% of
total cost for the first installation, including the cost of planning and
applications engineering. 2 Miller's estimates of the ratio of total
implementation cost to robot base price, and of the total cost of
implementing a single low-, medium-, and high-priced robot in
machine loading are shown in Table II. High-cost robots generally
have more degrees of freedom, a long reach, a large working
envelope, and more sophisticated control capabilities. Somewhat less
accessory hardware and engineering is required to enable such a
robot to acquire, load, and unload parts. Conversely, lower-cost
$160
~,\
Installation $24 \
(15%) \
\
\
\ $130 $130
\
\ \ /
\ \ /
Accessories $48 \ \ / $32 $26 $115
/
\ \ /
(e.g., Gripper (30%) \ \
$110 / (25%) (20%)
\ \ $18
,- ~
\ \ ~
$100 (16%)
\ \
\
\
\
\
/
--
/ 1/
/ $25
\ \ /
/
\ \ /
\ \ \ / (25%) / $26 $46 $46 $33
/
\ \ /
\ $75 ;' (27%)
\ \
/ (24%) (35%) (35%)
\
\
\ $8
\ $20 ~,
\ (U%) ,,"
\ (20%) ---
\ $17
\
\
Basic $88 \ (22%) ~~ $77
Robot (55%) (70%)
$52 $58 $64
(40%) (45%) (56%)
$50 $55
(67%) (55%)
Welding Material Machine Spray Assembly Machining All Robots

Handling Loading Painting
By Application ($000)
FIGURE 2
Typical robot system cost breakdown in selected applications. (From Tech Tran Corp., 1983, p. 97.)
20 STEVEN M. MILLER
TABLE II
Assumptions for Calculating Implementation Costs
Total implementation cost

Type of robot Base price Robot base price
Lower cost $20,000 4

Medium cost $60,000 3
High cost $100,000 2
robots are more limited in their degrees of freedom, reach, working

envelope, and control capabilities. Hence more accessory hardware
and engineering is required to compensate for these limitations. 3
Engelberger (1980, p. 102) has also emphasized this point.
Generally speaking the higher priced robots are capable of

more demanding jobs and their control sophistication assures
that they can be adapted to new jobs when original assignments
are completed. So too, the more expensive and more sophis-
ticated robot will ordinarily require less special tooling and
lower installation cost. Some of the pick and place robots are no
more than adjustable components of automation systems. One
popular model, for example, rarely contributes over 20% of the
total system cost.
Since 1979, there are many more models on the market as a

result of existing manufacturers (Cincinnati Milacron, Unimation,
TABLE III
Summary of Cost Assumptions for Retrofitting Level I Robot
Systems a
Robot Total Operators

hardware Development implementation replaced
cost (R) cost (D) cost (/ = R + D) per shift
20,000 80,000 100,000

60,000 180,000 240,000 1-2
100,000 200,000 300,000 1-3
a Source: Miller (1983., p. 44).

Prab, etc.) broadening their product lines, and as a result of many

new firms who have started to sell and manufactur~ robots over the
last five years (Bendix, GE, GM-Fanuc, IBM, and a number of
smaller firms.).4 One significant aspect of there being more models
available is that a user can shop around in the market until he finds a
robot with almost exactly the required bundle of attributes (e.g.,
reach, load, accuracy, type of control etc.) for the application in
mind. This means that a user need not pay for more capability than is
really necessary.
3. EXTENSIONS IN THE CAPABILITIES OF INSENSATE

ROBOTS
3.1. Increased Availability of High -Precision Robots for Light

Manufacturing
One important development has been a sharp increase in the
available robot models specifically designed to precisely position
lightweight workpieces. These robots are primarily intended for
small parts assembly, although they can be used in any application
where precise positioning of the end effector is required. Selected
characteristics of some of the earlier, as well as more recent,
high-precision robots are shown in Table IV. Such machines
generally have a repeatability in the range of 0.001-0.008 in. s Most of
these machines are designed to only handle lightweight payloads
within a relatively small working envelope. In contrast, robots with
longer reaches and with greater payload capacity, such as those used
for loading machines and press or for welding, typically have a
repeatability in the range of 0.01-0.08 in.
The machines made by Seiko Instruments, which have been in
use in Japan since 1960, were the first high-precision robots
developed for small parts assembly. They are extremely precise
because they use mechanical stops for positioning instead of servo
control. For the same reason, they cannot be reprogrammed without
adjusting the hardware. In 1973, ASEA introduced the IRb-6 model,
which was the first reprogrammable, servo-controlled robot with a
22 STEVEN M. MILLER
TABLE IV
High-Precision Robots Designed for Small Parts Assembly'
First
system
Approximate instal-
Payload Repeatability base price lation
Manufacturer Model (lb) (in.) ($000) date
Seiko 100 3 0.0004 6 1960

Seiko 400 9 0.002 10 1960
ASEA IRb-6 13 0.002 60 1973
Unimation PUMA 550\650 5 0.004 41 1978
General
Electric Allegro 14 0.001 125 1980
Unimation PUMA 260 2 0.002 41 1980
Westinghouse 4000 22 0.008 65 1980
American
Robot Merlin 50 0.001 60 1981
Automatix AID 600 17 0.003 85 1981
Bendix AA 45 0.002 70 1981
Yaskawa Motoman L3 7 0.004 60 1981
C. Milacron T 3 726 14 0.006 56 1982
IBM 7535 13 0.002 29 1982
IBM 7565 5 0.005 95 1982
Hitachi A3020 4 0.002 28 1983
a Sources: Compiled from Hunt (1983) and Tech Tran Corp. (1983).
repeatability approaching that of the Seiko. The repeatability of

0.002 in. was made possible in part by using an electric, as opposed to
a hydraulic, power system. In 1978, U nimation introduced the first
version of the PUMA, the first electrically powered servo-controlled
high-precision robot that can be controlled by downloading a
program from a computer as well as by the more conventional
method (teach pendant). Because it can easily be interfaced with a
computer, the PUMA quickly spread throughout industrial and
academic robotics research labs. In 1981, American Robot Corp.
introduced a computer-controlled robot that can position a 50-lb
payload with a repeatability of 0.001 in. This unusual combination of
high precision, a relatively large payload, and computer control are

the result of using a nonservo control system that is powered by
electric stepper motors. Several of the other recently introduced
models shown in the Table IV are licensed from robot producers in
Japan and Europe.
3.2. Increased Availability of Computer-Controlled Robots

Programmable robots first commercialized in the early 1960s,
such as the Unimate and the AMF Versatran (now Prab), had
computerlike functions, such as memory, but were made up of
special purposes electronic logic components designed to perform a
specific set of tasks. Electronic controls were only used to duplicate
the functions of other special purpose (hardwired) control functions.
Robots controlled by general-purpose digital computers were not
commercialized until the mid-1970s. The first minicomputer-
controlled robot was the r (566 and 568 models), commercialized in
1974 by Cincinnati Milacron. The first microprocessor-controlled
robot was developed in 1976 by VICARM, a small company founded
by inventor Victor Scheinman, who founded the company to build
small electric-powered arms primarily for research labs. Unimation
acquired the rights to the VICARM robot, further developed the
product, and commercially released it as the PUMA 500 robot in
1978. 6 To our knowledge, the r and the PUMA were the only
computer-controlled robots made by U.S. manufacturers com-
mercially available in 1979. Since that time, many other computer-
controlled robot models have appeared on the market. For example,
all of the high-precision robots introduced after 1978 listed in Table
IV are computer controlled.
The computer-controlled robot is far more flexible than a
machine controlled with specialized logic circuits. Because of power-
ful computational capabilities, it can be programmed in anyone of
several different coordinate systems. Robots without computer
control can only be programmed in one coordinate system. Because
instructions can be fed into the controller through a keyboard and
stored in computer memory, new robot programs can be developed
off-line, even while the robot is engaged in a task. In contrast,
24 STEVEN M. MILLER
"lead-through" and "walk-through" programming, the methods that

must be used if the robot is not computer controlled, are on-line
methods. The instructions are entered into the controller by physi-
cally positioning the robot arm and recording the position. To
develop new programs with on-line programming methods requires
that the robot be taken out of production.
Perhaps the most significant aspect of robots controlled by
general-purpose digital computers is that they can more easily and
more extensively communicate with other computers, and therefore,
with other machines controlled by computers. This has made it
possible to develop robot programs on a computer in one location (at
a centralized engineering facility) and to ship the program through a
network to the robot's computer on the floor of a factory in another
location. 7
The actions of a computer-controlled robot can be monitored
and supervised by an external host computer, which can also monitor
and supervise the actions of other machines. This capability has made
it possible to more extensively control and coordinate the interac-
tions of robots with other machines in manufacturing cells and
so-called manufacturing systems (FMS). In addition, computer-
controlled robots can more extensively process and utilize external
information provided from visual, force, and tactile sensors than
robots that are not computer controlled.
Experimental Improvements in Robotic Arms. Improvements in
the basic mechanical design of rigid robot arms are not highlighted in
this chapter. In fact, such research seems to have been deemphasized
lately.8 Nonetheless, there are noteworthy developments. Virtually
every robotic manipulator in use today is a rigid structure that uses
transmission mechanisms-gear trains, lead screws, steel belts, chains
and linkages-to transmit power from motors to the load. Physical
forces such as friction and backlash introduce unwanted disturbances
which cannot be completely eliminated, and performance capabilities
are constrained as a result. This limits the speed, accuracy, and
versatility of the arm. One significant development in the mechanics
of arm design is the direct drive manipulator developed at Carnegie-
Mellon University (Asada and Kanade, 1981). Lightweight but
high-powered direct drive electric motors are directly coupled to
each joint, eliminating transmission mechanisms between motors and

the load. Backlash is eliminated and frictional forces are minimized.
The behavior of this type of arm can be modeled more accurately
than an arm with conventional power transmission mechanisms.
More accurate modeling provides for improved control, enabling the
end effector to be moved at a higher speed and to be more precisely
positioned. Eliminating the transmission mechanisms also eliminated
mechanical parts, simplified maintenance requirements, and poten-
tially improved the reliability of the manipulator. The direct drive
arm paves the way for a new generation of lightweight, high-
performance robot arms. At least one manufacturer is experimenting
with a direct drive arm for commercial use.
3.3. The Integration of Sensory Information Processing with Robotic

Devices
The previous section focused on the use of internal sensing to
monitor the internal systems (e.g., joint positions, joint loads,
internal temperature, etc.) of a robot. This section focuses on the
sensing of external information, and on how this extends the
capabilities of the Level I robot. Nitzan (1981, p. 2) has given a
general definition of sensing:
We define sensing (regardless of whether it is human or

robotic, internal or external) in a broad way (i.e., as perception).
Sensing is the translation of relevant physical properties of
surface and volume elements into the information required for a
given application. The physical properties are electric, magnetic,
optical (e.g., surface reflectance and transmittance at different
wavelengths of the incident radiation), mechanical (e.g.,
presence/ absence, range, position, velocity, acceleration, stress,
and pressure), temperature, and so forth.
Figure 3 provides a framework for understanding how the

sensing and processing of external information extends the robot's
capabilities. Within the boxed region are elements representing the
robot control system (mechanisms for spatiotemporal coordination
and effector processing) and the end effector. These elements
26 STEVEN M. MILLER
, -------- ----- ,
,,, .---------, I
Spatio ,
System
Knowledge Temporal
Integration
Coordination
,,
Current Robots ;
r--~~--, ,
Sensory Effector
Processing Processing
,
- - - -,-
...... _- ... _-_ ....
FIGURE 3
Schematic diagram of a sensor-based robotic system. (From Sanderson, 1983, p. 33.)
comprise the insensate robot with open-loop control described

earlier. Outside the boxed region are elements representing devices
for sensing external information, computer programs for processing
and interpreting information (sensory processing and knowledge),
and means for communicating the results with the robot manipulator
(systems integration). With external sensing, knowledge can be
acquired from the external world and communicated back to the
robot controller. The position of the end effector can be measured
aQd adjusted using closed-loop control.
External sensing and closed-loop control are important capabili-
ties because they extend the range of manufacturing tasks that can be
accomplished by a robot. For some tasks it is not possible to
prespecify all motions to the level of exactness required to use an
insensate robot. For example, surfaces of many natural materials,
such as wood and stone, as well as surfaces of parts shaped in
metal-forming operations (e.g., sand casting, hammer forging) have
irregular and unpredictable variations which often have to be
smoothed over by grinding, routing, or deburring. In assembly
processes, often there are variations in the tolerances of parts which
have to be inserted into one another. Even tiny vanations can

complicate the insertion process significantly and require the use of
sensory feedback. Moreover, many tasks, as currently performed,
involve the selection of one or more small parts (washers, nuts, bolts,
resistors, etc.) from a bin or pile of randomly oriented parts. Finally,
an important function in manufacturing is to detect flaws in
components about to be assembled, or in completed parts or
subassemblies. External sensory information is inherently necessary
for all of the above tasks.
More specific reasons for why external sensing would extend the
range of robot applications in manufacturing are given by Rossol
(1983):
• Vision systems will provide positioning information to robots

so that parts, robots, and surrounding equipment need not be
precisely located.
• Vision systems will perform a range of inspection tasks
including incoming part inspection, in-process inspection, and
final quality control inspection.
• In a related area, force sensing and control systems will
provide fine positioning control, allowing for the detection of
unusual conditions such as collisions and make possible
high-precision assembly.
The use of external sensing might eventually enable preCISlOn

positioning to be achieved at a lower cost than by existing methods.
The total cost of an application would be reduced if the cost of
external sensing systems were to fall below the cost of mechanical
positioning devices. There is also the possibility of using less precise
robot manipulators and achieving precise positioning by measuring
and adjusting the final position of the end effector, much the way a
"floppy" human arm is positioned.
3.3.1. The Fundamental Problem of Machine Sensing. "Sens-
ing" is not merely the detection of signals reflecting physical
properties of the world. Referring back to Nitzan's definition at the
beginning of this chapter, sensing is the translation of relevant
physical properties of the world into information required for
28 STEVEN M. MILLER
carrying out a task. External data detected through the human

sensory organs (eyes, ears, hands) is automatically processed and
translated into the information required for a particular problem in
the brain. Because processing and interpretation is often unconscious,
humans typically do not distinguish between detecting signals from
the external world and interpreting what is detected. With machines,
however, it is essential to make this distinction. Computers must be
explicitly instructed on how to represent and store the physical
signals, process them, extract required information, and communicate
the information with the robot or other machines. Thus, for a robot
to "see" requires not only a TV camera, but sophisticated computer
programs to process and translate the electronic signals into useful
information. Detecting physical signals from the external world is a
relatively straightforward engineering task, and there are a great
variety of devices for doing this, as indicated in Table V. The primary
difficulty with machine sensing, however, is interpreting the signals
from the detectors and in translating these signals into useful
information.
TABLE V
Devices and Methods Available for Robots to Detect Signals
from the External World a
Visual detection devices Force detection devices
Photodetectors Probes
Linear arrays Strain gauges
Area arrays Piezoelectric materials
TV cameras Carbon materials
Lasers (triangulation) Discrete arrays
Lasers (time of flight) Integrated arrays
Optical fibers
Acoustic detection devices Other detection methods
Ultrasonic detectors/ emitters Infrared

Ultrasonic arrays Radar
Microphones Magnetic proximity
Ionizing radiation
a Source: Pugh (1983, p. 6).

Another issue is that the human brain is extremely flexible with

respect to the types of external information it can process and
interpret. It can apply different algorithms and appropriately filter
data according to the circumstances. At one instant, a tennis player or
a baseball hitter may concentrate totally on the approaching ball,
assessing its speed, direction, and spin. At a later instant, the athlete
may totally concentrate on the position of his opponent, on the
playing surface, or on the fans in the grandstand. An accountant can
scan a page of figures looking for the one or two numbers that are
not within the expected range. At one moment, an editor can skim a
page of print looking only for errors of punctuation and misspelled
words. At another moment, the editor reads the page for content. A
photographer may look at a landscape and concentrate on the overall
balance of light and shadow. A birdwatcher can select our small
moving creatures in the air, among clouds and trees. The same person
may be capable of all of these kinds of visual processing. In addition,
the person is also collecting and processing information received
through the hands and ears.
A machine, on the other hand, is extremely limited with respect
to the types of information that can be processed and interpreted. As
yet, there is no such thing as a general-purpose machine system that
could even remotely approach the full sensing capabilities of a
human. At first glance, the competition between the machine and the
human is highly unequal because the available sensing systems are
extremely crude in comparison to human capabilities. However,
many tasks do not require the full power of human sensing
capabilities (e.g., stereoscopic color vision). For extracting particular
specialized types of information from a scene, a machine sensing
system can actually be far more efficient than a human brain. In
general, machine senses can compete in effectiveness with human
sensing (disregarding cost for the moment) in one of two cases:
Case I. The machine is able to detect signals of a kind that are
not directly accessible to humans. For instance, the machine may be
able to directly sense voltages, currents, magnetic fields, frequencies,
amplitudes, or intensities or electromagnetic radiation outside the
optical range, pressures, temperatures, sounds, etc. Humans can only
detect such signals if they are electronically converted into analog or
digital readouts.
30 STEVEN M. MILLER
Case II. The machine is able to interpret a mass of data more

efficiently and faster than a human. For instance, computers can filter
and suppress much of the noise in a digitized satellite picture,
revealing salient features invisible to the unaided eye. Computers are
more efficient at matching fingerprints, because their speed enables
them to test quickly for a few key features. Computer vision systems
can be designed to efficiently identify the edge of objects In
two-dimensional patterns and measure the length and width of
objects faster and more accurately than humans.
Case I situations are prime candidates for substitution of
computer sensing systems for human workers. The reason is that
external sensors are required in any event to convert the physical
signal into a form that can be comprehended by a human. All that is
needed to eliminate the human operator from the control loop is a
means of inputing the detected signal into a microprocessor (e.g., an
analog-to-digital converter). 9 This substitution has already occurred
in computer numerical control machines because the basic "state-of-
the-system" data for such machines is in a form more accessible to
machine sensors than to human eyes, ears, and hands. The vast
majority of the uses of external sensing systems, especially machine
vision systems, fall into Case II. Here, we are talking about using
external sensing systems as direct substitutes for human sensing.
3.3.2. Increases in the Use of Machine Vision. Pugh (1983, p. 4)
comments that "if it is perhaps vision more than the senses of touch
and hearing which has attracted the greatest research effort." For this
reason, our overview of developments in the use of external sensing
concentrates on the use of machine vision. Kanade (1983) describes
the general process of vision as follows:
Vision involves visual sensing and interpretation. Visual

sensing is a projection of a physical environment into a form of
representation called images. Projection can vary from the most
ordinary picture taken by cameras to active sensing such as by
laser rangefinders. Images can also range from a single-point
light flux measurement to 3-dimensional range data. Then vision
is defined as a process of understanding the environment
through the projected images, or in short inverse projection. 10
Typically, when referring to machine vision, we mean a two-

dimensional pattern, characterized by varying degrees of intensity
projected from a three-dimensional scene. Visual information is often
used to inspect critical features of objects, to recognize and identify
objects, and to locate the position of objects in space.
3.4. Commercial Use of Machine Vision Systems Since 1919

From several indicators, it is clear that there has been a large
increase in the application of machine vision systems in industry
since 1979. As of the end of 1982, the were 400-500 machine vision
systems reportedly installed in U.S. manufacturing operations, com-
pared to only a handful of installations reported in 1979 (R.
Sanderson, 1983). Whereas at most a few firms supplied vision
systems in 1979, 22 manufacturers of machine vision systems and
over 20 university or government research institutes with machine
vision programs (outside of industrial R&D) in the U.S. alone are
listed in R. Sanderson (1983). Over 128 manufacturers, developers,
and distributors of vision systems, image-processing systems, and
components worldwide are listed in Zimmerman, Van Boven, and
Oosterlinck (1983).
One important factor contributing to the increased usage in
vision systems since 1979 was the commercialization of the system
developed at SRI (Agin, 1980). The SRI system provided the basic
product technology for many manufacturers of vision systems,
including Machine Intelligence Corp. and Automatix. Recent ap-
plications of industrial applications of vision systems can be found in
Zimmerman and Oosterlinck (1983), in the proceedings of the
International Conference on Robot Vision and Sensory Controls
(1979 through present) and also in the proceedings of the Interna-
tional Symposium on Industrial Robots, especially the more recent
ones (e.g., the 11th, 12th, and 13th Symposiums).
Capabilities of vision systems commercially available and in
research labs are compared in Table VI based on Bolles (1981). The
general problem-solving approach of commercially available systems
is to simplify the scene to be analyzed, thereby simplifying the
information processing requirements. Research systems are designed
32 STEVEN M. MILLER
TABLE VI
Comparison of the Capabilities of Commercially Available Vision Systems
and Research Vision Systems·
Attribute Current commercially

of system available systems Research systems
General problem- Extensive use of task Use of more sophisticated

solving approach specific engineering vision systems that can
to simplify the scene function under less
to be analyzed structured conditions
Properties of Constrained viewpoint Unconstrained viewpoint
scenes that Isolated parts Partially constrained content
can be analyzed Known set of parts Arbitrary set of parts
Rigid parts Nonrigid parts
Lighting Very carefully controlled: Less carefully controlled:
Uniform illumination Nonuniform illumination
over field of view over field of view
High contrast Low ,contrast
No shadows Shadows
No specular reflectance Specular reflectance
No photometric distortions Distortions
Fixed position and Varying position and
intensity of light sources intensity of light sources
Data Often binary High spatial and
intensity resolution
Occasionally multifrequency
data
Occasionally range data
Representation Usually just one type Often several redundant
of image types and levels
Use of parametrized
object models
Knowledge Compiled into program Combination of
compiled and explicit
knowledge
Strategy Extract features and Requires search
make decisions based on and some reasoning
feature classification and
table look-up
Computers Small and inexpensive Large and expensive
a Source: Adapted from Bolles (1981).

to operate under less structured conditions and therefore require

more sophisticated information processing. A partial list of factors
affecting the degree of difficulty of processing visual information
includes the properties of the scene, the lighting, the camera-object
relationship, the range of spectral data processed, the way in which
images are represented within the computer, the way in which
knowledge is used for problem solving, and the strategy used to
make decisions about the image. For each of these factors, the table
indicates in the first column the restricted conditions required for
most commercial systems to function. The less restricted conditions
under which research systems can function are indicated in the
second column.
4. FLEXIBLE MANUFACTURING SYSTEMS
4.1. An Overview of Manufacturing Systems

We broadly interpret the term "manufacturing system" to mean
the following:
• a set of tools for processing materials;

• a means for moving materials from one tool to another;
• a means for controlling and monitoring the action of the tools
and the movement of materials.
Various types of manufacturing systems are distinguished in

Table VII, based on general attributes of the processing tools,
material handling system, and means of supervisory control and
monitoring. Of the automated systems in the table (the first four
types), the one with the longest history of use is the specialized
transfer line. Such systems may be very large, including up to several
hundred interconnected process tools. Each tool in the system is
single purpose, highly specialized to perform one type of task.
Material flow from one process tool to another is automated, but the
sequence of the flow is fixed, predetermined in advance, and cannot
be altered without stopping production and reconfiguring the system.
34 STEVEN M. MILLER
TABLE VII
An Overview of Manufacturing Systems
Specialized Semiflexible Stand

transfer transfer Flexible alone
Attributes line line system Cell tool
Process tools:
Number of Up to 3-100 3-50 1-10 One
interconnected several
tools hundred
Degree of Single- Multi- Multi- Multi- Multi-
specialization purpose purpose purpose purpose purpose
of each tool
Handling of
material flows:
Across tools Automated, Automated, Automated, Automated, None
in network fixed fixed variable variable
sequence sequence sequence sequence
Into and Automated Automated Automated Automated None
out of
each tool
Supervisory control:
Of material Dedicated Programmable Hierarchical Hierarchical Manual
flows logic logic computer computer
control control control control
Of tool Same Same Same Same Manual or
parameters computer
The loading and unloading of material into and out of each process
tool is also automated. Control of the movement of materials from
one process tool to another, as well as the control of the parameters
of each tool, is implemented via dedicated logic controllers built in
(or hardwired) into the machine hardware.
A semiflexible transfer line differs from a specialized one in that
some or all of the process tools in the interconnected line may be
multipurpose, capable of performing a range of tasks. Such systems
may include many process tools, but typically not as many as used in
the specialized transfer lines. Although the sequence of the flow of
materials from one tool to another is still fixed, the sequence of
operations at each tool is not, since several operations can be carried

out at each multipurpose tool. This added flexibility at each tool
makes it possible for a semiflexible transfer line to process a wider
range of parts than a specialized transfer line. A limited degree of
programmability is required to control the timing of the material
flow and the parameters of each tool. Typically, a semiflexible
transfer line is controlled by the programmable logic controllers
(PLC).
A flexible manufacturing system (FMS) differs from a semiflex~
ible transfer line in that neither sequencing of the flow of materials
from one tool to another nor the sequence of operations at each tool
is fixed. The material handling system has "random bypassing
capability," i.e., a part can be moved from anyone tool in the
interconnected system to another because the transport system can
bypass any tool along its path, on demand. Thus, each part can
traverse a variable route through the system. This flexibility in
material handling, in combination with the use of multipurpose tools,
makes it possible for a flexible manufacturing system to process a
greater diversity of parts than a semiflexibe transfer line. Because of
the larger number of possibilities, control of automated material
flows and of tool parameters in this type of system is considerably
more complex than in a semiflexible transfer line. Typically, a flexible
manufacturing system requires a hierarchical control system where a
computer supervises and coordinates the actions of the control
mechanisms in the material handling system and in the process tools.
Such systems typically have fewer process tools than transfer lines,
partly because the requirements for extensive information processing
and control limits the number of elements that can be supervised and
coordinated.
An outline of a typical hierarchical control scheme used in a
flexible manufacturing system for machining is shown in Figure 4.
The type of parts to be produced, volume requirements, and due
dates are inputed into the FMS supervisory computer by the system
manager, which is typically a human but which may be another
computer. Information on when and how each part is to be processed
flows from the supervisory computer at the top level of the hierarchy
to the material handling system and machine tools at the lower level.
The control mechanisms within the material handling system move
36 STEVEN M. MILLER
Parts to be machined Inputted by hUman}

Processing and volume requirements per part { supervisor or by
factory computer
•
Supervision of part movements Supervision of machine tools
Determination of Determination of
- which parts to process - which NC program
- when to start each part to ship to each
- which pallet to use for Reporting on the machine
each part start and finish - which NC program
- which machine to use for of each NC should be used
each processing step program by each machine
- where to move each pallet at a particu lar
- when to move each pallet time
Reporting on the - when to start a
Reporting on status of each given NC program
the location machine and of at each machine
and status of each tool
each part and
each pallet
Material Handling System Individual NC Machine Tool
- control material handling - Local storage of NC programs

system to move pallets between - control machine tool parameters
machine tool tables, load/unload (axes of motion, power drives,
stations and storage stations, tool change sequence, etc.),
given commands from supervisory given commands to NC program
program - Control movement of parts back
and forth from pallet to
machine tool table
FIGURE 4
Overview of hierarchical control system III a flexible manufacturing system for
machining.
the palletized parts around in such a way that the higher-level

commands of the supervisory computer are satisfied. The material
handling system "reports" back information on where parts are, and
whether they have been machined or not, to the supervisory
computer. The supervisory computer also controls and monitors
some actions of the machine tools. The supervisory program

determines which NC programs should be residing at each machine
tool, and transfers the appropriate program to the appropriate tool
when necessary.
The difference between a manufacturing cell and a flexible
manufacturing system is more one of degree than of kind. A system
with multipurpose tools, variable sequencing material handling, and a
computer supervising the control of the material handling system and
tool parameters would be called a cell if there were only three or four
different tools in the system. If there were a larger number of tools,
say five or ten or more, it would be called a flexible manufacturing
system or FMS. There is no clear-cut boundary between large cells
and small flexible systems. The control scheme used in a manufactur-
ing cell would be essentially the same as that shown in Figure 4.
However, the material handling "system" may now only consist of
one or several robots. A cell could even be considered as a single
"stand-alone" work station. Within the cell, the flow of materials is
automated, but the flow of materials into and out of the cell is
handled manually. A flexible manufacturing system could be con-
sidered as a network of manufacturing cells connected via a material
handling system. The flow of materials is automated within each cell,
as well as across all cells within the network.
The simplest example of a manufacturing system in this
framework would be a single stand-alone process tool. Typically, the
machine would be multipurpose. There is no automated flow of
materials from machine to machine, since each machine is not
interconnected with other any except by manual transfer. The
loading and unloading of material into and out of the machine is
manual. If the loading and unloading were automated, the stand-
alone tool would in effect become a cell. Automatic control is not
required to supervise a manual material flow system. While there is
usually an operator, some aspects of the machine tool parameters
may be taken over by a computer. This is the least sophisticated
manufacturing system from the viewpoint of the use of automation,
but is the most adaptable. Engineering changes are easily incor-
porated and the time required to change tools is minimal.
Kearney & Trecker, Corp., a pioneer in the manufacture of
38 STEVEN M. MILLER
t i... 8
High ;:3
~
000-
Special System
000
iun:
Flexible Manufacturing
500 System
25 Manufacturing Cell
Stand Alone N/C
Low
Low 1 or 2
.
MedIum
J
8 100
Part Numbers Per System
FIGURE 5
Kearney & Trecker volume/variety chart. (From Kearney & Trecker, 1982.)
FMS, has developed a volume/variety chart (Figure 5) to illustrate

where it is appropriate to use the various types of manufacturing
systems for machining applications. The term "volume" refers to the
number of parts per year manufactured with the system. "Variety"
refers to the number of distinct part designs run through the system.
The specialized transfer line is appropriate when production
requirements lie within the top left quadrant of Figure 5, the region
of highest volumes per part and only a few different part types. There
is almost no limit on the maximum number of identical parts per year
a transfer line may be built to produce. There are also special cases in
which a line may also be used to produce as few as 2000 pieces of one
part number per year. Stand-alone tools are most appropriate when
production requirements fall within the lower right quadrant of the
figure, the region of lowest volume per part and a large number of
different part types. In machining applications, this typically cor-
responds to 300 or more part different types per year, with most
parts produced in batches of 50 or less. Prior to the development of
intermediate forms of manufacturing systems in the mid-1950s, the

only alternative choices were the two extremes of the spectrum.
The three more recent variants of automated manufacturing
systems, semiflexible transfer lines (labeled as special systems in
Figure 5), flexible manufacturing systems, and manufacturing cells,
are designed for midvolume and mid variety production requirements.
In machining applications, this typically includes yearly production
volumes in the range of 15-15,000 units per part combined with
varieties ranging from 2 to 800 different part types. Within this range,
semiflexible transfer lines are designed for large batch production. In
machining applications, this roughly corresponds to volume require-
ments ranging from 1000 to 15,000 units per part, and part varieties
ranging from 2 to 10 different part types. Flexible manufacturing
systems are designed for medium-sized batch production, typically
including volume requirements ranging from 50 to 2000 units per
part and part varieties ranging from 4 to 100 part types. Manufactur-
ing cells are designed for small batch production, typically including
volumes ranging from 10 to 500 units per part and varieties in the
range of 10-800 different part types.
4.2. Flexible Manufacturing Systems for Machining
Flexible manufacturing systems for the machining of parts have

been in commercial use for over 15 years in the United States. The
first systems were built by White-Sundstrand and Kearney &
Trecker. A partial list of FMS currently in use in the U.S. is given in
Table VIII. More detailed descriptions of FMS for machining are
given in Jablonowski (1980), Kearney & Trecker (1982), Dupont-
Gatelmand (1983), Draper Lab (1983), and Manufacturing Engineer-
ing (1983). Klahorst (1983a), of Kearney & Trecker Corp., reports
that most FMS systems will produce between 5000 and 75,000 parts
per year. He reports that attempts to justify an FMS for fewer than
5000 workpieces will usually fail because of the resultant high cost of
automation per workpiece produced. Justification of an FMS for
more than 75,000 parts will usually fail because the system will be a
more costly alternative than more specialized forms of automation
(transfer lines),u Total production volumes of the systems listed in
40 STEVEN M. MILLER
TABLE VIII
Partial List of EMS for Machining in Use in the United States·
Voluml and
Date number of Types of
User Vendor installed partl parts
Sundstrand White- 1967 nad Aluminium

Aviation Sundstrand na pump parts
Ingesoll- White- 1970 20,000b, 14C Hoist and
Rand Sundstrand motor cases
Rockwell K&T 1972 25,000b, 45 C Automotive
axle carriers
Allis- K&T 1970, 1973 23,000b, 8c Agricultural
Chalmers equipment
Caterpillar White- 1973 6600b, 6C Crank case
Sundstrand housings, covers
AVCO- K&T 1975, 1978 24,000b,9 c Aircraft
Williamsport engines
AVCO- K&T 1979 15,000b, 10C Turbine
Lycoming engines
Caterpillar Giddings 1980 na Construction
and Lewis na equipment
John Deere K&T 1981 SOOOb, 8c Farm
equipment
International Giddings 1981 na
Harvester and Lewis na
General Giddings 1981 5600b,7c Motor
Electric and Lewis housings
AVCO Giddings 1982 na na
and Lewis na
Detroit White- 1983 na,40 c Transmission
Diesel Allison Sundstrand housings
• Source: Draper Labs (1983).

b Total number of parts per year machined on system.
c Number of different part types machined on system.
dna: information not available.
Table VIII range between 5000 and 25,000 parts. Total production
volumes of the systems built by Kearney & Trecker range between
5000 and 60,000 parts (Klahorst 1983a).
Available information on investment in FMS hardware and
reported savings in operating costs are shown in Tables IX-XII.
TABLE IX
Hardware Investment and Reported Savings in Operating Costs for Selected
Kearney & Trecker Machining FMS·
Product volume
Cost and Reported comparisons
User (1982$) part variety with old system
Rockwell $5.6 million 24,OOOb, 45 c 114 floor space;

(truck axles) setup costs
virtually eliminated
AVCO- $8.4 million 24,OOOb, 9c 113 floor space;
Williamsport 114 labor;
(aircraft engines) 112 number of
part holding devices;
John Deere $18 million 50,OOOb, 8_12c Cost estimates of
(tractor components) FMS: $18 million;
dedicated transfer line:
$28 million;
Mack Truck $5 million 65,OOOb, 5C Cost estimate of FMS
(truck components) about the same as
estimate for dedicated
transfer line, with
comparable cycle time,
less flexibility;
Caterpillar d $5 million 8000 b, 8C Total transit time
(construction equip.) through system:
Old system: 8.5 h
New system: 0.3 h
a Source: Klahorst (1983b).

b Total number of parts per year machined on system.
c Number of different types machined on system.
a This system is a semiflexible transfer line. All other systems listed here are flexible manufacturing
systems. See Table VII for definitions.
Investment in system hardware runs into the multimillion dollar

range. Users report large savings in operating costs, including
reductions in labor cost, reductions in cost associated with in-process
inventories, and reductions in cost associated with floor space
requirements.
The hardware costs of the Kearney & Trecker systems listed in
Table IX range between $5 and $18 million. This sum includes the
cost of the machine tools, material handling equipment, control
42 STEVEN M. MILLER
TABLE X
Hardware Investment and Reported Savings in Operating
Costs for Machining FMS at General Electric, Erie a
Old New
machining machining
Comparison system system
Number of machines 29 9
Total production 86 16
worker requirements
for two shifts
(operators, supervisors, and
maintenance)
Typical number of 10-11 4-5
machine loadings required to
complete one pan
Maximum annual output 4100 5600
for family of
seven pans
Average in-process time 16 days 16h
for a part
Cost Base $16 million
Productivity Base +240%
change (total factor)
a Source: Miller (1983b).
systems, tools and fixtures, and testing, training manuals, and other
support services. This sum does not include the costs of planning,
engineering, and installation. There are no available data on these
"support" costs. Kearney & Trecker has found that it typically takes
three years of planning between the time a customer decides to buy
an FMS and the time a system is installed by a vendorY Considering
that almost every functional area of manufacturing is affected by the
use of the FMS and must be involved to some extent,13 it seems that
the costs of planning and engineering are considerable. The FMS
users listed in Table IX report a variety of savings in operating costs
including substantial reductions in the amount of labor, tooling, and
floor space required to produce a given level of output. In two cases
in the table, the price quote for the FMS was equal to or less than the
TABLE XI
Hardware Investment and Reported Savings in Operating Costs for
Matching FMS System at Yamazaki Machinery Works a
Plant Yamazaki Machinery Works Ltd., Nagoya, Japan
Process: Machining
Products: Parts for numerically controlled machine tools and robots
Total costs: $20 million
Total employment: 13
Comparisons: A conventional machining system with similar production
volume would require 215 workers and nearly four times as
many machines. Average in-process time per part was
reduced from three months to three days. Over a five-year
period, the plant is expected to be 15 times more profitable
than a conventional plant its size.
a Source: Bylinski (1983).
TABLE XII
Hardware Investment and Reported Savings m Operating Costs for
Machining FMS at Fanuc Ltd:
Plant Fanuc Ltd., Fuji complex, Japan
Process: Machining
Products: Parts for machine tools and robots
Total cost: $32 million
Comments on hardware: 30 machining centers, consisting of computer-controlled
machine tools loaded and unloaded by robots, along
with material-handling robots, monitors, and
programmable controllers.
Employment: About 100
Comparison: The plant is about five times more productive than its
conventional counterpart. It would probably take 10
times the capital investment and 10 times the labor force
to produce the same output with a conventional plant.
a Source: Bylinski (1983).

44 STEVEN M. MILLER
price quote for a special purpose transfer line. Thus, in some

production situations, FMS can actually reduce the investment
required to automate, in addition to providing the advantages of
flexibility.
A comparison .of machining locomotive parts (motor frame
housings) with a conventional system of stand-alone machines and
with a new FMS at General Electric's Erie facility is shown in Table
X. The FMS has one third as many machine tools as the system it
replaced. Twenty-nine manually operated machines were replaced by
nine automated machining centers. As a result, floor space require-
ments were reduced by 25% and the typical number of times a part
had to be loaded onto a separate machine was cut in half. The total
number of people required to support the machining activity over
two shifts (material handlers, operators, maintenance workers sche-
dulers, and supervisors), was reduced from 86 to 16. The capacity of
the FMS is 38% greater than the capacity of the manually operated
stand-alone tools. The new system is designed to produce 5600 parts
per year, whereas the old system produced about 4100 parts per year.
GE management claims that capacity of the FMS could be increased
by an additional 60%, to 9000 parts per year, with only minor
modifications of the current system. The average in-process time was
reduced from 16 days to 16 h. The hardware cost of the system was
about $16 million. GE management reports a 240% increase in total
factor productivity as a result of installing the new system, although
they do not specify over what time period this increase has been (or
will be) realized.14
Two large Japanese flexible manufacturing systems are described
in Bylinski (1983). An overview of the FMS at the Yamazaki
Machinery Works, in Nagoya, Japan, is given in Table XI. Yamazaki
management claims that a conventional system with manually
operated stand-alone machine tools would require nearly 15 times as
many workers and four times as many machines. They calculate that
over time, the new FMS will be 15 times more profitable than a
conventional plant its size. The company estimates that over five
years of operation the plant will produce after-tax profits of $12
million compared with $800,000 for a conventional plant that size.
The hardware cost of the system is estimated to be $20 million. 15 An
overview of the FMS at Fanuc, Ltd., in Fuji, Japan, is given in Table

XII. Fanuc management reports that a conventional system would
require ten times the number of workers and capital investment to
produce the same output as the new FMS. They claim that the plant
is about five times more productive than its conventional counterpart
would be. The hardware cost are estimated at $32 million. 16
Yoshikawa, Rathmill, and Hatvany (1981, pp. 11-12) provide a more
detailed description of the Fanuc plant:
Recently, there has been a remarkable trend toward having

systems unmanned during the night shifts. Fanuc has con-
structed a new factory in Fuji that does this. This plant has 29
cell-like work stations. Seven of them are equipped with robots;
22 are equipped with automatic pallet changers with pallet pools.
These stations are connected by unmanned vehicles guided by
electromagnetic or optical methods. The plant has two automatic
warehouses, one for materials and another for finished parts and
subassemblies. The vehicles transport the materials from the
warehouse to the unmanned machining stations. Robots or
automatic pallet changers load materials onto the stations from
the vehicles. Finished parts are transferred again automatically
by the vehicles to the second warehouse.
The plant has an assembly floor, where workers work
during the day. Transportation between the parts warehouse and
this floor is also by unmanned vehicles. In the daytime, 19
workers are working around machining stations, mainly for
palletizing, and 63 workers are in the assembly section. Thus, 82
workers are in the plant during the day, but at night there is
only one. The assembly floor is closed, and the machining floor
is operated without any workers. Every station is equipped with
a monitoring device with a TV camera, and one person sits in the
control room to monitor all the working stations. He observes
the working status of stations through the camera, without
touring the factory. The monitoring device also records the
spindle motor current, calculating the cutting force and time to
judge cutting conditions.
Insensate (Level I) robotic manipulators are sometimes used in

the large FMS to load and unload parts, as indicated above. One
46 STEVEN M. MILLER
robot is used in the GE system to move cutting tools into and out of
a machine tool. In general, robotic manipulators do not playa major
role in these types of systems. The primary reason is that the basic
transformation task, the machining, is already automated by the
machine tool. Most of the other tasks are related to material handling
and there are often alternative forms of automation to move parts and
tools. For example, automated guided vehicles are often used to
transport parts from the loading station to the machine tool. There
are also many ways of moving a tool from a tool rack to the tool
storage compartment in the machine tool. Robots are used more
extensively in manufacturing cells, which are smaller than the
large-scale FMS. If the machines in the cell are properly arranged, it
is typically less expensive to use one or several robots to move
materials from one machine to another than to use other types of
material handling equipment.
An analogy can be made between the first generation of flexible
manufacturing systems used for machining and the Level I industrial
robot. Like the Level I robot, the flexible manufacturing systems
described above are essentially insensate. 17 As a result, much of the
cost of the complete system is a result of having to engineer the
domain to eliminate all possible sources of unexpected variation out
of the system.
Both the Yamazaki and Fanuc plants make parts for machine
tools and robots. Given the examples of cost savings cited earlier, one
would expect a reduction in the cost of machine tool and robot
components manufactured on flexible manufacturing systems. A
reduction in the cost of components would have a substantial impact
on total robot cost, given that for widely used models, the cost of
components comprises three quarters of the total manufacturing cost
(Engelberger, 1983). Thus, while robots do not playa major role in
the large FMS for machining, these systems will eventually have a
substantial impact on the cost (and therefore on the use) of robots.
4.3. Developments in Flexible Manufacturing Systems

Three developments in flexible manufacturing systems are dis-
cussed in this section:
• The integration of multiple FMS for machining in one plant;

• The use of FMS in manufacturing processes other than

machining;
• The addition of artificial intelligence and sensor-based systems
to FMS.
In the flexible manufacturing systems described in Tables

IX-XII, one FMS was installed in the factory. The largest of the
systems includes about 30 machine tools, while many of the systems
include fewer than 20. Using available technology, it is possible to
build systems of flexible manufacturing systems with larger number
of workstations. Mazak Corp., the American affiliate of Yamazaki,
has recently built a factory in Florence, Kentucky with five flexible
manufacturing systems for machining. Each system produces a
different family of parts. In total, there are 60 computer numerically
controlled machines in the five systems, and a central supervisory
system can communicate with all 60 machines. The hardware cost of
the plant, which includes assembly as well as machining, is estimated
at over $200 million. This does not include planning and engineering
costs (Klotz, 1984).18
Flexible manufacturing systems are now being tested and
marketed for applications other than machining. U nimation, a
subsidiary of Westinghouse, is marketing flexible manufacturing
systems for sheet metal fabrication and for wire harness assembly.
Researchers at the Carnegie-Mellon Robotics Institute have
developed a FMS for open-die forging (Wright et at., 1982) and for
electronics assembly (Sanderson and Perry, 1983). Both systems have
been transferred to industry. Numerous examples of flexible manu-
facturing systems for assembly are found in the literature. Westing-
house developed an FMS to assemble small electric motors, but the
system was only used for experimental purposes (Stauffer, 1983).
IBM has developed an FMS to assemble display writers (Stauffer,
1982). Other flexible manufacturing systems for assembly are dis-
cussed in Nevins and Whitney (1979), Warnecke and Walther (1983),
Kanesaka (1983), and Freeman (1983). A hybrid flexible assembly
system, using both robots and humans, is discussed in Lien (1983).
All of these systems use robotic manipulators for parts transfer. All
of the assembly systems also use robots to perform the basic
assembly tasks.
48 STEVEN M. MILLER
Within the next several years, FMS will be used in a wide range
of manufacturing processes. The next logical step would be to
integrate the flow of material from one FMS to another. Early
attempts to integrate material flow across processes in flexibly
automated systems have already been made. In the Fanuc plant,
automated guided vehicles move material from an automated
warehouse to the machining FMS. In the Maze plant, the automatic
carts used to transport parts through the machining FMS are also
used to move parts to and from manual assembly areas.
The flexible manufacturing systems for open-die forging and
electronics assembly developed at the CMU Robotics Institute make
extensive use of new technologies; computer programming techni-
ques that emerged from the discipline of artificial intelligence (AI)
and external sensing. The open-die forging system is controlled by a
new rule-based computer language (Bourne, 1982) that is based on
the AI programming technique of production systems. One advan-
tage of this rule-based language is that it simplifies the process of
designing a control system that is robust and resilient to unanti-
cipated errors in the system. Conventional programming languages
(FORTRAN or PASCAL or APT) are procedural. There is a sequential flow
of control from one statement in the program to the next. At any
random statement in the program, the proposed task to be performed
is critically dependent on the previous task performed. If one task is
not carried out because of a machine or program error, a subsequent
task may be adversely affected or not enacted. If the control system is
written in a procedural language, it is difficult to prevent a break in
any link of the chain from breaking the whole control system. With a
rule-based system, which is nonprocedural, the flow of control is
determined by detecting which conditions hold and then by execut-
ing rules that specify what to do given the particular set of
conditions. This is in contrast to following a predetermined sequence.
If a machine or part of the program fails, the control program will
detect that a new set of conditions now holds, and then execute the
rules that govern what to do under those conditions.
A second advantage of rule-based systems is that they can easily
accommodate changes in the physical system. Additions of new
equipment do not require major revisions of the control program as
they would with control systems written with conventional pro-

gramming languages. A third advantage is that the rule-based
structure makes it possible to control unwanted interactions, such as
two robots moving to the same place at the same time. Once
programs written in conventional procedural languages get very
large, it is nearly impossible to anticipate all possible system
interactions. Events such as collisions could happen under a freak set
of circumstances. The rule-based language makes it possible to add
rules that preclude certain interactions from ever happening, even if
the reasons giving rise to these interactions are unanticipated. This AI
approach to controlling a system, which is robust to failure, easily
modified and extended, and protective against unanticipated harmful
interactions among system components, is one reason why an FMS
could be applied to this new application. The FMS described in
Wright et al. (1982) also incorporates a machine vision system which
is used to inspect the shape of the part that emerges from the
open-die forge.
The electronics assembly cell described in Sanderson and Perry
(1983) makes extensive use of external sensing systems to detect
visual and force information. In addition, the system makes extensive
use of techniques that have emerged from AI to process and interpret
the sensory information. Vision is used to guide a robotic mani-
pulator in acquiring an unoriented electronic component. Both visual
and force sensing are used to guide the manipulator in inserting
electronic components into a printed wire board. These technologies
make it possible to build an FMS for the assembly of small electronic
components. Also, the sensor-based technologies make it possible to
combine processes, such as assembly and inspection, into one system.
Applications of sensor-based systems and AI are also being
developed to improve the capabilities of the flexible manufacturing
systems for machining. Vendors are experimenting with machine
vision systems to verify if the correct part is being processed on a
particular machine, to inspect parts after they have been machined,
and to inspect tools to see if they need replacement. Also, vision is
being tested to eliminate the need for precisely positioning each part
and each tool in the machining system. 19
These examples indicate that the range of processing tasks that
50 STEVEN M. MILLER
can be integrated into a flexible manufacturing system will expand as

the technologies of AI and external sensing improve and become part
of FMS technology. There is also indication that these technologies
will improve the performance of flexible manufacturing systems in
areas where they are already being used.
ACKNOWLEDGMENT
This is a condensed version of the paper "Recent Developments

in Robotics and Flexible Manufacturing Systems," prepared as part
of a technology assessment of industrial robots supported by the
National Science Foundation, Division of Policy Research and
Analysis, grant No. PRA-8302137-826. I am especially grateful to
Robert Ayres, principal investigator of the technology assessment
project, for ideas and for editorial assistance and to Thomas Morton
for reviewing the manuscript.
NOTES
1. Eikonix Corp. (1979), pages 26-27.

2. Miller notes that the average cost per application decreases as more
robots are installed in a factory, especially when there are multiple
installations of a similar application (see Miller, 1983a, Chap. 2.) If more
than one robot were installed, the cost of robots would comprise a
smaller proportion of the total installation cost.
3. There are exceptional cases where the limited capabilities of the low-cost
robot are well suited to the task and a minimum of accessory hardware is
required.
4. Engelberger asserts that as of late 1983, there are approximately 230
companies in the world selling robots.
5. Repeatability is a measure of a robot's ability to return to the same
position over many repetitions. A repeatability of 0.005 in. means that
the robot will return to within 0.005 in. of its original position on each
successive replication. See Albertson (1983) for a more extensive
discussion of repeatability.
6. The 500 model had five degrees of freedom. The 600 model had six
degrees of freedom. The model numbers were later changed to the

PUMA 500 and 650.
7. For example, in the robotic assembly system reported in Sanderson and
Perry (1983) robot programs are developed on a Vax 11/750 and then
transmitted to a smaller computer, a PDP 11123, and then finally
transmitted to the controllers of the individual robots. The Vax is also
networked to other computers in the research environment.
8. According to Engelberger, the technology underlying virtually all
insensate robots in use today is the servo control technology developed
during World War II. He believes there will be no major breakthroughs
in the robot arm itself as long as robots are built as stiff structures with
the positioning based on classical open loop servo control theory. Given
this constraint, he argues that emphasis should be placed on redesigning
(or designing) the product for ease of manufacturing, and on integrating
robots and other types of machinery into manufacturing systems. A
major development in robot arm technology, in his opinion, requires the
development of a lightweight flaccid arm that is precisely positioned by
sensing the position of the end effector and the target position
(Engelberger, 1983).
9. This is discussed in more detail in Ayres (1984).
10. Page 59.
11. Klahorst (1983a), p. 53.
12. Kearney & Trecker (1982), p. 14.6.
13. Kearney & Trecker (1982), p. 14.6.
14. Performance comparisons were obtained directly from GE personnel
during a visit to the facility (Miller, 1983b). These comparisons are
corroborated in Manufacturing Engineering (1983), pp. 66-67, and in
Bylinski (1983).
15. Tom Klotz, sales manager of Mazak (the U.S. affiliate of Yamazaki)
claims that the FMS also reduced floor space requirements by over half,
from 70,000 square feet to 30,000 square feet. Klotz also reports the
following benefits in another Yamazaki plant (The Seiko Works).
Number of machines: reduced from 54 to 24. Number of workers (three
shift total): reduced from 105 to 14. Floor space: reduced by one third.
In process time: reduced from 13 weeks to three days (Klotz, 1984).
16. While visiting the United States in the summer of 1983, Professor Yukio
Hasegawa, an experienced Japanese robotics researcher from Wasada
University in Japan, corroborated the information on the two Japanese
systems reported in Bylinski (1983).
17. These systems may use bar code technology to identify which part has
52 STEVEN M. MILLER
been placed on a pallet. These systems do not yet use machine vision or
taction or other forms of external sensing to detect the position and
orientation of parts or to inspect parts.
18. Klotz estimates the design required hundreds of thousands of engineer-
ing hours.
19. These experimental applications of machine vision systems in machining
FMS were provided by representatives from Cincinnati Milacron and
Mazak.
REFERENCES
Agin, G. J. (1980). Computer vision systems for industrial inspection and

assembly, Computer (IEEE Computer Society) 13, 11-20.
Albertson, P. (1983). Verifying robot performance, Robotics Today 5(5),
33-36.
Asada, H. and Kanade, T. (1981). Design of Direct-Drive Mechanical Arms,
Technical Report CMU-RI-TR-81-1, Carnegie-Mellon University Ro-
botics Institute.
Ayres, R. U. (1984). The Sensory Feedback Dependence of Assembly Tasks
as a Measure of Comparative Advantage of Human Workers vis-a.-vis
Sensor-Based Robots, Technical Report, Engineering and Public Policy
Department, Carnegie-Mellon University.
Bolles, R. C. (1981). An Overview of Applications of Image Understanding
to Industrial Automation, Technical Note 242, SRI International.
Bourne, D. A. (1982). A Numberless, Tensed Language for Action Oriented
Tasks, Technical Report CMU-RI-TR-82-12, Carnegie-Mellon Univer-
sity Robotics Institute.
Bylinski, G. (1982). The race to the automatic factory, Fortune 107(4),
52-64.
Coiffet, P. (1983). Robot Technology, Vol. 1: Modelling and Control,
Prentice-Hall, Englewood Cliffs, New Jersey.
Draper Lab Staff (1983). Flexible Manufacturing System Handbook, Vol. II:
Description of the Technology, Technical Report 12703, Charles Stark
Draper Lab, prepared for the US Army Tank-Automotive Command,
Warren, Michigan.
Dupont-Gatelmand, C. (1983). A survey of flexible manufacturing systems,
]. Manuf. Sys. 1(1), 1-16.
Eikonix Corp. (1979). Technology Assessment: The Impact of Robots (Final
Report), Technical Report ECI2405801-FR-1, Eikonix Corp., NSF

grant No. ERS-76-00637.
Engelberger, J. F. (1980). Robotics in Practice, American Management
Association, New York.
Engelberger, J. F. (1983). Private Communication, Interview with Steve
Miller, Danbury, Connecticut.
Freeman, G. T. (1983). P.C.B. Automation in the 1980's, in 15th CIRP
Seminar on Manufacturing Systems: Assembly Automation, pp. 329-334,
International Institute for Production Engineering Research (CIRP).
Hunt, V. D. (1983). Industrial Robotics Handbook, Industrial Press, New
York.
Hunt, H. A. and Hunt, T. L. (1983). Human Resource Implications of
Robotics, W. E. Upjohn Institute for Employment Research, Kalama-
zoo, Michigan.
Jablonowski, J. (1980). Aiming for flexibility in manufacturing systems, Am.
Mach. 124(3), 167-182.
Kanade, T. (1983). Visual sensing and understanding: The image understand-
ing point of view, Compo Mech. Eng. 1(4), 59-70.
Kanesaka, S. (1983). Flexible Micro Coil Assembly System CAS-20, in 15th
CIRP Seminar on Manufacturing Systems: Assembly Automation, pp.
308-328, International Institute for Production Engineering Research
(CIRP).
Kearney & Trecker Corp. (1982). Special Products Division. Manufacturing
Systems Applications Workbook. Kearney & Trecker Corp., Milwaukee,
Wisconsin.
Klahorst, H. T. (1983a). How to plan your FMS. Manuf. Eng. 91(3),52-54.
Klahorst, T. (1983b). Interview with Steve Miller. Private research notes on
visit to Kearney & Trecker Corp.
Klotz, T. (1984). The Flexible Manufacturing Factory, Presentation made at
the Flexible Manufacturing Systems Seminar sponsored by the Robot
Institute of America, January 26, 1984, Pittsburgh, Pennsylvania.
Lien, T. K. (1983). Flexible Control of an Integrated Automatic-Manual
Assembly Line, Manufacturing Systems: Proceedings of the CIRP
Seminars 12(3), 218-226.
Luh, J. Y. S. (1983). An anatomy of industrial robots and their controls, in
Tutorial on Robotics, (Lee, C. S. G., Gonzalez, R. c., and Fu, K. S.
eds.), pp. <;-25, IEEE Computer Society Press, New York.
Manufacturing Engineering Magazine (1983). Special issue on flexible manu-
facturing systems, M anuf. Eng. 91 (3).
54 STEVEN M. MILLER
Miller, S. M. (1983a). Potential Impacts of Robotics on Manufacturing Costs

in the Metalworking Industries, Ph.D. dissertation, Carnegie-Mellon
University.
Miller, S. (1983b). The GE Erie Facility, Private research notes on visit to the
GE Erie facility.
Nevins, J. L., and Whitney, D. E. (1979). Robot assmebly research and its
future applications, in Computer Vision and Sensor-Based Robots,
(Dodd, G. D. and Rossol, L. eds.), pp. 275-321, Plenum, New York.
Nitzan, D. (1981). Assessment of robotics sensors, in Proceedings of the 1st
International Conference on Robot Vision and Sensory Control, (Pugh,
A. ed.), pp. 1-11, IFS Publications, London, England.
Paul, R. P. (1981). The MIT Press Series in Artificial Intelligence: Robot
Manipulators: Mathematics, Programming and Control, MIT Press,
Cambridge, Massachusetts.
Pugh, A. (1983). Second generation robotics, in International Trends in
Manufacturing Technology, Vol. 1: Robot Vision, (Pugh A. ed.), pp.
3-10, IFS Publications, London.
Rossol, L. (1983). Computer vision in industry, in International Trends in
Manufacturing Technology, Vol. 1: Robot Vision, (Pugh, A. ed.), pp.
11-18, IFS Publications, London.
Sanderson, A. C. (1983). Robot vision and industrial automation, in
Industrial Applications of Image Analysis, (Zimmerman, N., and
Oosterlinck, A. eds.), D.E.B. Publications, Pijnacker, The Netherlands.
Sanderson, A. C. and Perry, G. (1983). Sensor-based robotic assembly
systems: research and applications in electronic manufacturing, Proc.
IEEE 71(7), 856-871.
Sanderson, A. c., and Weiss, L. E. (1983). Adaptive visual servo control of
robots, in International Trends in Manufacturing Technology, Vol. 1:
Robot Vision, (Pugh, A. ed.), pp. 107-116, IFS Publications, London.
Sanderson, R. J. (1983). Machine Vision Systems: A Summary and Forecast,
Tech Tran Corporation, Naperville, Illinois.
Stauffer, R. N. (1982). IBM advances robotics assembly in building a word
processor, Robotics Today 4(5), 19-23.
Stauffer, R. N. (1983). Westinghouse advances the art of assembly, Robotics
Today 5(1), 33-36.
Tech Tran Corp. (1983). Industrial Robots: A Summary and Forecast, Tech
Tran Corp., Naperville, Illinois.
Warnecke, H. J., and Walther, J. (1983). "Programmable Assembly Station
for the Automatic Assembly of Car Aggregates," in 15th CIRP Seminar
on Manufacturing Systems: Assembly Automation, pp. 290-299, Inter-
national Institute for Production Engineering Research (CIRP).
Wright, P. K., Bourne, D. A., Colyer, J. P., Isasi,]. A. E., and Schatz, G. S.
(1982). A flexible manufacturing cell for swaging, Mech. Eng. Mag.
October, 76-83.
Yoshikowa, H., Rothmill, K., and Hatvany, J. (1981). Computer-Aided
Manufacturing: An International Comparison, Technical Report, Na-
tional Academy Press, sponsored by the National Research Council.
Zimmerman, N., and Oosterlinck A. (eds.) (1983). Industrial applications of
image analysis, D.E.B. Publishers, Pijnacker, The Netherlands.
Zimmerman, N. J., Van Boven, G. J. R., and Oosterlinck, A. (1983).
Overview of industrial vision systems, in Zimmerman, N., and Ooster-
linck, A. (eds.), Industrial Applications of Image Analysis, Chap. 10, pp.
203-229. D.E.B. Publications, Pijnacker, The Netherlands.
CHAPTER 2
PERSPECTIVES ON
AUTOMOTNE MANUFACTURING
Steven A. Weiner
1. INTRODUCTION
This chapter provides a description along with a personal

perspective on some aspects of automobile manufacturing at Ford
Motor Company. In particular, the chapter begins with a brief
overview of the scope of Ford's North American manufacturing
operations. After this overview, certain aspects of vehicle assembly
and stamping operations are discussed in somewhat greater detail.
The next subject to be covered is scheduling of manufacturing
operations, both at the level of the assembly plant (first tier) and at
the stamping plant as representative of a second tier supplier. With
this material as background, new operating strategies in manufactur-
ing such as computer integrated manufacturing (elM), just-in-time
aIT), flexible machining, and statistical process control (SPC) will be
considered from the perspective of change required for implementa-
tion. Finally, the chapter will conclude with a short summary.
Throughout this overview, the focus will be on opportunities and/or
requirements for change in order to achieve greater productivity and
improved product quality.
Steven A. Weiner. Manufacturing Processes Laboratory, Ford Motor Company,

24500 Glendale Avenue, Detroit, Michigan 48239.
57
58 STEVEN A. WEINER
2. SCOPE OF FORD NORTH AMERICAN

MANUFACTURING OPERATIONS
As of January, 1982, Ford's North American Automotive

Operations (NAAO) utilized over 100 million square feet of floor
space in 87 major facilities, exclusive of Ford's integrated glass, steel,
casting, electronic, climate control, plastics, paint and vinyl, and
aerospace operations. The latter operations are quite sizable in their
own right, utilizing over 39 million square feet in 73 major facilities.
Ford's wholly owned steel subsidiary, Rouge Steel, is the ninth
largest steel company in the U.S., and its Glass Division is the largest
maker of the float glass. In 1982, Ford-U.S. and domestic sub-
sidiaries' average employment was 155,900 people.
Ford's North American manufacturing operations include a
wide spectrum of processes and technologies. A broad understanding
of the character of these plants will provide background for the
remainder of this chapter. To simplify the description of these plants,
they have been grouped into four categories (Figure 1). In the most
basic of the operations, raw materials such as iron ore and sand are
converted to the engineering materials used in the vehicles. Some
plants take these materials and manufacture individual components
such as body sheet metal stampings and chassis forgings. These
operations consist of many complex mechanical, chemical, and
thermal processes and include a large base of processing experience
No. of Percent Capital/Energy Labar

Plants In - House Intensity Intensity
Raw Materials
11 60 High Low
I I
Conversion
26
j
Component Mfg.
Finishing €A
13
Sub- Assembly
Vehicle Assembly 19 100 Low High

FIGURE 1
Scope of Ford North American manufacturing operations.
PERSPECTIVES ON AUTOMOTIVE MANUFACTURING 59
and expertise. The next level of operations combines certain com-

ponents into the major subassemblies, such as transmissions, engines,
rear axles, doors, and body frames. They are characterized primarily
by machining and welding/joining processes, component assembly,
and by extensive part handling and inspection. The vehicle assembly
operations consist of joining the subassemblies, adding trim, finishing
processes, and complex material handling and inspection procedures.
In order to bound the discussion, this chapter focusses on the vehicle
assembly operations, and on stamping operations which are repres-
entative of the first level of suppliers to the vehicle assembly plants
from a scheduling perspective.
3. VEHICLE ASSEMBLY OPERATIONS
The typical Ford North American automotive assembly plant

consists of five basic shops or areas: body, paint, trim, chassis, and
predelivery. In North America, the vehicle assembly plants are
supplied with stampings from separate plants which typically supply
more than one assembly plant. In Ford European Operations, the
stamping plant is an integral part of the body shop. In either case,
body shop operations are highly automated and consist mainly of
welding together of sheet metal components to form the body-in-
white, i.e., an unpainted body. The typical body shop will have over
95% of the welds made automatically using a combination of both
robots and fixed automation. The body shop is the prime determin-
ant of vehicle fit, i.e., the mating of the components so as to eliminate
gaps and to maintain vehicle appearance. In North America the cost
of transporting stampings to the vehicle assembly plant is a sig-
nificant factor in the design of stamped components and in the
decision as to whether to partially assemble stampings at the
stamping plant prior to shipment or at the assembly plant. By way Of
illustration, such considerations led to the conclusion that the Escort
door should be assembled at the stamping plant prior to shipment.
Also, the North American Escort'~ uses a multiple-piece bodyside to
* Throughoutthis chapter, the term "Escort" is used to include the Escort, Lynx,
LN7, and EXP car lines.
60 STEVEN A. WEINER
form the door opening, the pieces being shipped separately to the
assembly plant, where highly specialized tooling is used to achieve
the precise door opening.
After the body-in-white is completed, it is moved into the paint
shop. There the unit is washed, phosphated, and painted. The
activities in the paint shop are intended to provide corrosion
protection along with an aesthetic finish. Like the body shop, the
paint shop is highly automated. However, unlike the body shop, the
paint shop cannot be simply stopped in place or shut down at
breaktime. It also cannot be as readily restarted. This is because in
the paint shop there are many operations such as washing, phosphat-
ing, spray painting, and baking where the vehicle must be taken
completely through the operation before it can be left "in place."
This is so even though, like the body shop, the paint shop is highly
automated.
After the body-in-white has been painted, it is now ready to be
processed through the trim and chassis shops for completion of
vehicle build, including installation of the interior and the
powertrain. Trim and chassis shop operations have an inherently high
requirement for manual labor, although for certain aspects, such as
engine installation, there are automation assists. It is also in these two
shops that the impact of the marketing decision to allow multiple
customer options is mainly felt. Using the Escort as an example,
there are approximately 2.5 million potentially unique configurations
including color. However there are only 10 body styles, e.g., 2-door,
4-door, station wagon, etc. and less than 20 exterior paint treatments.
The impact of option complexity (large numbers of options) is slight
in the body shop where each body style has its own unique fixture
and! or stamped component. In the paint shop, complexity leads to
an increase in the amount of material used to purge the paint spray
guns on color changeover. However, in the trim and chassis shops,
the customer decision to order such options as a stereo radio or a rear
window defroster changes the labor content of the vehicle and leads
to constraints on vehicle sequencing for the trim and chassis shops.
This resequencing of vehicles prior to the first trim shop operation is
done using a painted body bank. These banks typically consist of
4-10 parallel conveyors, with each having a capacity of from 7 to 20
painted bodies. To illustrate the operation of such a bank, consider an

example in which the bodies leaving the paint shop contain a string of
three vehicles in a row (\1;, \1z, \1;) which call for a rear window
defroster. Furthermore, let us suppose that the trim shop is balanced
(manned) so that it can install a rear window defroster every fourth
unit. Thus, these three units should arrive at trim in a string of nine
vehicles positioned first, fifth, and ninth (\1;, X, X, X, \1z, X, X, X,
\1;). In order to achieve this sequence (or rotation) in this simple
example, the first vehicle would be assigned vehicle rotation number
1, the second vehicle rotation number 5, and the third vehicle
rotation number 9. These vehicles could then be stored in the painted
body bank in a single file in the order 1, 5, 9. The next vehicles
arriving from the paint shop (assuming no rear window defroster)
would then be assigned rotation numbers 2, 3, 4, and 6, 7, 8,
respectively. In this case, it is assumed that the painted body bank
contained sufficient units so that the trim shop could be supplied
with vehicles while units 1, 5, 9 were held pending arrival of the six
additional bodies. Once the nine vehicles were assigned the complete
rotation sequence 1-9, the descriptions of these vehicles would then
be broadcast to the trim and chassis shops, including off-line
operations, so that the necessary engines, seats, radios, and other
options can be assembled in the right order.
At the end of the chassis shop, the vehicle is drivable. Work in
predelivery consists of final vehicle inspection, headlight alignment,
and such minor repairs as might be required. The typical assembly
plant produces finished vehicles at the rate of 30-80 jobs per hour
aPH) depending upon manning levels, vehicle work content, and the
degree of automation. In support of this final rate, the upstream
shops tend to have a 5-10% higher job per hour capacity with the
body shop having greater capacity than the paint shop followed by
the trim and chassis shops, respectively. This arrangement allows the
final rate to be maintained barring major production stoppages. In
the main, assembly plant capacity is adjusted by changing the number
of hours worked rather than by adjusting the line rates. The
particular layout also places great stress upon first run capability in
the paint shop. In particular, if the paint shop goes down for any
length of time the painted body bank empties and resequencing is no
62 STEVEN A. WEINER
longer possible. Under this scenario the line balance constraints of

the trim and chassis shops will be exceeded on a dynamic basis. This
leads to a requirement for additional manpower in order to avoid a
deterioration in product quality. Using our example, if the painted
body bank is empty when vehicles \1;, Vz, and \-j arrive, the vehicles
will be sent down the trim line in that sequence. Clearly, additional
labor will be required to install three rear window defrosters in a row
when the line is manned for no more than one in four. However,
even if the additional labor were available, the regular operator would
still be preferred because of the greater experience in performing the
task.
4. STAMPING OPERATIONS
Ford's 19 North American assembly plants are fed by six

stamping plants, four of which have identical layouts. The material
flow of one of these four plants is shown schematically in Figure 2.
The plant is on a 411 acre site and covers roughly 2.7 million square
feet. At capacity, some 200-250 end items could be produced,
consisting of doors, hoods, fenders, deck lids, and other major
stampings from 1100 tons of steel per day.
Steel coils arrive at the area shown at the left of the figure and
are then loaded onto blankers or shears. There, the coils are cut up to
the rough size of the desired stampings. The blanks then move
through the press lines where the stampings are formed. Depending
upon the part, the stampings are shipped directly to an assembly
plant or are assembled into a component prior to shipment. For
example, the door assembly undergoes ten manufacturing operations
prior to shipment.
ASSEIoIBLY
LINE
BLANKS
FIGURE 2
Schematic of stamping plant material flow.
On closer examination, two major process steps occur in the

blanking area: flexrolling, and blanking and shearing. Approximately
40% of incoming steel is flexrolled prior to stamping. This process is
used to enhance the formability of outer panels. However, it carries
with it the burden that flexrolled steel blanks must be used within 24
hours to avoid age hardening.
There are 13 blanking presses and 11 shears which operate at
speeds up to 60 strokes per minute, or 3600 pieces per hour. Other
sources of complexity are the multiple gauges, widths, and types of
steels used as well as incoming material variability. The magnitude of
the latter problem is illustrated in Figure 3 for incoming steel to be
used for stamping the Escort door outer panel. The percent of
incoming steel rejected per month for the first 22 months after the
start of production is plotted as a histogram. Note the skewness of
the distribution and the occurrence of the points corresponding to
31 % and 47% rejection. The mean rejection rate was approximately
10%, with a standard deviation of 11 %.
After blanking, the stacks of steel blanks are then moved by fork
truck to the press lines. Some of the press lines are completely
automated, some utilize the side sliding bolster presses (which are
currently receiving much publicity), while others use front to back
sliding bolster presses. In total, there are 26 major press lines and 11
Me an = 9.8%
4 r- Mo de = 7%
St d.Dev. = tt.4
(/)
w 3
u
z
w
a:
a: 2 r- r-
::>
u
u
o
- r-
o
o 5 10 15 20
PERCENT REJECTED PER MONTH
,
(
~J,L
FIGURE 3
Quality of incoming steel for Escort door outer panels.
64 STEVEN A. WEINER
intermediate lines. A major press line typically has 4-6 stations with
automated part handling between stations. The process steps are
drawing or forming, trimming, flanging, and cropping. Up to seven
different parts are run on the same press line. Captive lines are used
for high volume parts. Typical line rates are 550-750 pieces/hour.
The typical die running period is three days. When die changes are
required, current practice is to change the dies during scheduled
down shifts.
In the assembly area, the most common process technology is
welding. Other processes are a mechanical hemming process, mecha-
nical stitching, and adhesive bonding. There are approximately 80
dedicated subassembly areas and assembly lines to produce the
assembled end items, or one area/item. Purchased parts are also
incorporated into final assemblies. The final assemblies are then
loaded into special storage racks for shipment. As would be expected
there is a large pressline/ assembly area buffer.
One interesting and important aspect is the inherent mismatch of
process capabilities in stamping and vehicle assembly operations. To
be specific, as noted earlier, vehicle assembly plants typically operate
at 30-80 JPH. The stamping plant assembly (welding) lines operate at
200 JPH and are in turn fed by press lines which operate at
550-750 JPH for major panels, which in turn are fed by blankers
which operate at 1000 JPH. Clearly the scheduling of such inherently
different operations is a complex problem. This problem is further
exacerbated by the decision to minimize in-process inventory and by
inherent uncertainties in transportation (assuming separate plant
locations). Furthermore, the problem of incoming material quality,
both at the stamping plant and at the vehicle assembly plant, must be
dealt with if inventories are to be reduced while production schedules
are tightened.
5. SCHEDULING
Having described the nature of both vehicle assembly and

stamping plant operations, let us now turn to the problem of
short-term production scheduling. To begin with, all vehicles are
66 STEVEN A. WEINER
dividual actions on overall profitability. These problems are exacer-

bated by the fact that the lower tier stamping plant must make the
requested components or else the assembly plants are shut down.
Furthermore, as we move to zero inventory (exclusive of pipeline
inventory) the supplier facilities, in reality, must manufacture to
order rather than just ship to order.
The magnitude of the scheduling instability is illustrated by the
data shown in Figures 4 and 5. In Figure 4, the weekly orders for the
combined total of Escort, Lynx, EXP, and LN7 front left-hand doors
are shown for the period from February 7, 1982 to July 24, 1982. As
might be imagined, the instability is worse at the individual part level.
Thus the comparable demand for only the Escort front left-hand
door is even more erratic, as shown in Figure 5. Clearly, it is very
difficult to run a plant at all, much less efficiently, faced with such
fluctuations in demand.
In addition to demand instability, there is also the instability
brought about by such factors as production upsets including
30
(3
oo 25
(J)
20
o0::
o
o
Q;
X 15
w
~
0::
8
(J)
10
W
...J
j:! 5
g
WEEK
FIGURE 4
Total Escort/Lynx/EXP/LN7 doors demanded weekly for the period from February
7,1982 to July 24,1982.
built to order. The district sales offices consolidate orders and

transmit the information to the scheduling office, where it is in turn
broken down into 12-day order banks for each of the assembly
plants. At the individual assembly plant level, the order bank is then
processed to yield weekly and daily breakdowns. The processing is
first done to support production constraints in. terms of option
content and fixed capacity limitations. For example, only a certain
percentage of vehicles in anyone day can be station wagons owing to
the limitations on the number of body side welding fixtures. In
addition to these facility constraints, which tend to be constant, there
are also varying constraints due to material availability. This is a
particularly serious problem where, for example, major powertrain
components are sourced overseas. Typically, the pipeline is so long
and the scheduling and production practices so constrained that it is
not possible to dynamically match the supply to demand variations.
Rather, the supply is relatively constant and the item is either
restricted (and vehicle orders not selected for build) or inventoried.
These practices create a wave in building such vehicles and can lead
to high premium freight charges. Another consequence of the varying
material constraints is that it is rare that the vehicle mix is stable over
a period of days or weeks. In fact, given a production rate of roughly
1000 vehicles/day, typically at least 800 of them would be unique.
For the assembly plants, the last step in the scheduling system is to
sequence this daily bucket, i.e., the 1000 vehicles, to support the
body and paint shop priorities.
In parallel with this activity, the daily buckets from all the
assembly plants are submitted to the scheduling office, which then
consolidates the build schedules and generates an order tape for the
assembly plants' suppliers. In essence, this tape consists of one week
of "firm" orders as perceived by the supplier plus three weeks of
anticipated orders for planning purposes.
Once this tape is received at the stamping plant, a daily schedule
for the week's firm orders is then generated manually. Among other
factors taken into consideration are the status of machines and dies,
inventory counts, and actual plant manning levels.
The major problems with this system stem from the lack of
scheduling stability and the inability to assess the impact of in-
14400
12800
11200
o
Z 9600
<[
~
~ 8000
...J
<[ 6400
:::)
tJ
<[ 4800
3200
1600
1. .21 .23
2/14/82
WEEK
FIGURE 5
Total left-hand Escort carline only doors demanded weekly for the period from
February 7, 1982 to July 24, 1982.
manpower instability and by transportation-related causes. Taken

together, these factors lead to a high degree of manual intervention in
scheduling systems. Even with computer-assisted scheduling, the
dominant scheduling criterion is feasibility. For the complex multi-
stage production settings encountered in automobile manufacture,
scheduling theory is not sufficiently developed to be directly
applicable (Parker and Rardin, 1982). Nonetheless there are several
heuristic procedures which if realistically applied show potential for
improved plant scheduling (Kan 1976; Weiner et al. 1983).
6. "NEW" OPERATING STRATEGIES IN AUTOMOTIVE

MANUFACTURING
With the reemergence of manufacturing as a vital element in

profitability has come a renewed interest in "new" operating
68 STEVEN A. WEINER
strategies (Andersen and Co., 1983; Schonberger, 1983; Monden,

1983; Sadowski, 1984, Deming, 1981; Janotik and Johnson, 1984;
Dupont-Gatelmand, 1982). In this section, brief comment is made on
just-in-time OIT) inventory policy (Anhur Andersen & Co., 1983;
Schonberger, 1983; Monden, 1983), statistical process control
(SPC) (Denning, 1981), improved machining processes 0 anotik and
Johnson, 1984; Dupont-Gatelmand, 1982), and computer inte-
grated manufacturing (CIM) (Monden, 1983; Sadowski, 1984).
In a recently released Delphi survey, Arthur Andersen &
Company surveyed vehicle manufacturers' and parts suppliers' views
on JIT (Arthur Andersen & Co, 1983). Both agreed that the largest
problem to be overcome in implementing JIT is the unstable vehicle
assembly schedule. Again, a quick review of Figures 4 and 5
illustrates this point. Recognize that since the demand plotted is for
the Escort left-hand door it is also implicitly a plot of the number of
Escorts demanded each week. It is also recognized that JIT will not
enhance overall North American automotive industry productivity if
all it does is to shift inventory from manufacturer to parts supplier
(Arthur Andersen & Co., 1983; Schonberger, 1983). From the manufac-
turers' perspective, the other two major problems that need to be
overcome are the need for consistent, defect-free quality and the
obtaining of the understanding and commitment of top management.
While the latter is also the third concern of parts suppliers, their
second concern is the communication and cooperation of vehicle
manufacturers. Notwithstanding these concerns, both groups agree
that JIT will be a fact of life before the end of the 1980s.
In parallel with the move to JIT, North American automobile
manufacturers are also revitalizing their programs in statistical
process control (SPC) (Deming, 1981). Often this is being done in
conjunction with Employee Involvement or Quality of Work Life
programs. These programs are being used to improve process perfor-
mance and to identify obstacles to further improvement. There is also
an ongoing campaign to implement SPC methodology at parts
suppliers sites in conjunction with JIT implementation. The goal is to
improve process reliability and product quality, simultaneously.
In addition to improvements in SPC, there are also im-
provements being made in relatively mature process technologies
such as machining. One such advance is high metal rate removal

(HMRR) machining Oanotik and Johnson, 1984). The HMRR
machining concept is evolutionary in nature and is based upon a
stru.ctured selection procedure. Productivity is improved by employ-
ing higher cutter speeds and feeds, up to 30 ml sec and 1 mml
revolution, respectively. The machines themselves must be different
and more dynamically stable. These machines run at constant
rotational speeds and have improved power utilization. The specific
optimal cutting conditions are selected by computer algorithms
backed up by extensive machine and tool databases. By way of
illustrating representative productivity improvements achievable via
HMRR machining, a turning operation on a gray iron cylindrical
part is performed in production at a rate of 56 JPH. Under HMRR
conditions, this part was machined in the laboratory at a rate of
135 JPH for a productivity gain of 140% Oanotik and Johnson,
1984).
Another potential advance in automotive manufacturing is the
development of flexible manufacturing systems (Dupont-Gatelmand,
1982). Conventional automotive transfer lines are designed to make
only one or two parts at annual volumes in excess of 100,000 parts.
Traditional flexible manufacturing systems are designed to machine
6-20 parts at annual volumes of approximately 1000 pieces (Dupont-
Gatelmand, 1982). There are many automotive machining applica-
tions in which the part family consists of a handful of parts with
annual volumes in the 10,000-20,000 part range. Currently, the
special systems available in this range are either relatively slow or
capital intensive. There are also potential automotive applications for
part families of less than 10 parts with annual volumes in the
500,000-1,000,000 part range.
Lastly, in the area of "new" manufacturing strategies, is
computer integrated manufacturing (CIM) (Sadowski, 1984). Here,
the key is to use computer-based information strategies to improve
the overall manufacturing flow. Generally, such a notion underlies
most descriptions of future factories. In particular, CIM is regarded
as the cornerstone to unmanned operation. Yet, like JIT, CIM must
also overcome problems associated with demand instability and
incoming material quality. In addition, implementation of CIM will
70 STEVEN A. WEINER
require adoption of wholly new support systems along with the

training and! or retraining of manufacturing personnel in computer
based techniques. One example of a support system which must be
replaced under CIM is the accounting system used to calculate
manufacturing costs. Most manufacturing accounting systems are
based upon direct labor costs (Bornholtz, 1982). Often accounting
analysis will consist of analysis of "critical ratios" such as the ratio of
indirect to direct labor. Clearly, such analyses become meaningless as
the amount of direct labor tends to zero. Another practice, the
allocation of maintenance resources based on direct labor hours
worked, would also appear to be counterproductive in a CIM
facility.
In order to implement these "new" operating strategies in
automotive manufacturing, training and! or retraining of current
personnel in the use of modern manufacturing techniques is required.
Like most American manufacturers, Ford uses a seniority system for
hourly personnel and for salaried personnel, all other things being
equal. At some of our plants, seniority dates are now back to the
mid-1960s. Thus, for many, formal training!retraining will be
required in basic engineering areas such as applied mathematics,
statistics, and problem analyses as well as in the use of modern
technology. There are grounds for optimism, in part, based on the
experience in implementing CAD (computer-aided design) techni-
ques, that experienced personnel are eager and able to learn these new
methods. For this corporate commitment to traininglretraining to
succeed, educational institutions may have to make or renew
commitments to provide technological training to mature individuals.
Course content and course requirements may need revision to reflect
the greater practical experience and the lesser knowledge of modern
theory. For manufacturing engineering to become both a useful and
an accepted academic discipline, academic theory and industrial
practice need to be combined.
1. SUMMARY
In summary, the current difficult environment in the automotive

industry has led to the reemergence of manufacturing as a vital
element in corporate profitability. New competitors, coupled with

unstable demand and long lead times, have combined to reveal the
weaknesses in current production practices. On the positive side,
advances in manufacturing technology and in operations management
strategies show promise in restoring North American automotive
manufacturing productivity. Training and/or retraining personnel in
the use of advanced manufacturing technologies and mechanized
decision support systems, including mechanized scheduling systems,
is already in progress as part of a vigorous campaign to enlist the full
participation of all automotive industry workers in productivity and
quality improvement. The combination of employee participation
with new high-technology manufacturing methods will enable North
American automotive manufacturers to better cope with the changes
in objectives, resources, and constraints which characterize the
automotive industry of the 1980s.
ACKNOWLEDGMENT
I would like to thank A. M. Janotik for supplying the example

used to characterize HMRR machining and D. D. Dodge for
obtaining and analyzing the data on quality of incoming steel. I also
thank the latter and D. G. Lee for reviewing this manuscript. Lastly,
I would like to acknowledge my colleagues at Ford: B. T. Bajorek,
K. W. Casey, P. E. Coffman, J. W. Grant, S. E. Hoffman, J. W.
Knight, S. B. O'Reilly, C. D. Piercecchi, S. L. Pohly, and L. Y. Tsui,
for sharing the insights gained in their analyses of Ford manufactur-
ing operations.
REFERENCES
Arthur Andersen & Co. (1983). The Changing U.S. Automotive Industry,
Arthur Andersen & Co., Tinley Park, Illinois.
Bornholtz, E. F. (1982). Company and cost accounting, in Handbook of
Industrial Engineering (G. Salvendy, ed.), pp. 9.1.1-9.1.20, Wiley, New
York.
Deming, W. E. (1981). Management of Statistical Techniques for Quality and
Productivity, draft copyright W. E. Deming, Washington, D.C.
72 STEVEN A. WEINER
Dupont-Gatelmand, C. (1982). A survey of flexible manufacturing systems,"

]. Manuf. Syst. 1, 1-16.
Janotik, A. M., and Johnson, J. S. (1984). Application of Silicon Nitride
Tools to High Metal Removal Rate Machining, SAE Technical Paper
840532, Society of Automotive Engineers, International Congress and
Exposition, Detroit, Michigan, February 27- March 2, 1984.
Rinnooy Kan, A. H. (1976). Machine Scheduling Problems, Martinus
Nijhoff, The Hague.
Monden, Y. (1982). Toyota Production System, Industrial Engineering and
Management Press, Norcross, Georgia.
Parker, R. G., and Rardin, R. L. (1982). An overview of complexity theory
in discrete optimization: Parts I and II, IIE Trans. 14,3-10,83-88.
Sadowski, R. P. (1984). Computer integrated manufacturing, Ind. Eng. 16,
35-40.
Schonberger, R. J. (1983).]apanese Manufacturing Techniques: Nine Hidden
Lessons in Simplicity, Industrial Engineering and Management Press,
Norcross, Georgia.
Weiner, S. A., Grant, J. W., Pohly, S. L., Tsui, L. Y., and O'Reilly, S. B.
(1983). On the Use of Simulation Models to Evaluate Alternative
Schedules, Winter Simulation Conference, Arlington, Virginia, Vol. 1, p.
125.
CHAPTER 3
IMPACTS OF ROBOTICS AND

FLEXIBLE MANUFACTURING
TECHNOLOGIES ON
MANUFACTURING COSTS
AND EMPLOYMENT
Steven M. Miller
1. INTRODUCTION
The issues analyzed in this paper are the extent to which unit
costs and production labor requirements might be reduced in
manufacturing industries if there is more widespread use of industrial
robots and flexible systems. The analysis is reported in detail in
Miller (1983). These issues are analyzed from two different perspec-
tives. The technological focus of the first perspective is narrowly
confined to the use of robotic manipulators. It is assumed that
robotic manipulators will be "retrofitted" into existing production
facilities without making major changes in the organization of
production within the factory, other than modifying individual work
stations so that robots can replace one (or perhaps several) operators.
The critical variable in this perspective is an estimate of the percent of
the production worker jobs that will be replaced by robots.
Reductions in unit cost are calculated by assuming that a given
Steven M. Miller. Graduate School of Industrial Administration, Carnegie-Mellon

University, Pittsburgh, Pennsylvania 15213.
13
74 STEVEN M. MILLER
percentage of labor costs is reduced. Cases are also considered where

robot use results in a moderate increase in output as well as a
decrease in production labor requirements. The question of whether
decreases in production labor requirements could be offset by an
increase in demand stimulated by a reduction in price is also
analyzed.
The technological focus of the second perspective is much
broader than the first, and is concerned with the impacts of
integrating robots with other types of computer-assisted manufactur-
ing (CAM) technologies into flexible manufacturing systems. It is
assumed that a factory using general-purpose machines to produce
specialized products in batches can be reorganized and integrated so
that machines are fully utilized and used more efficiently. One critical
variable in this perspective is an estimate of the potential increase in
output that could be realized if all of the time in a year available for
production were utilized. The other critical variable is an estimate of
the unit cost and of the labor requirements in a fully utilized batch
production plant. Based on an analysis of a large cross section of
metalworking industries, a relationship is specified between the level
of output and the level of unit cost. Reductions in unit cost for a
given increase in output are derived from this relationship. Reduc-
tions in unit labor requirements are calculated in a similar manner.
2. METAL WORKING INDUSTRIES
To date, 80%-90% of the robots used in the United States and

in Japan, as well as in the rest of the world, have been installed within
a subset of manufacturing industries referred to as the metalworking
sector. For this reason, the analysis of the potential impacts of robot
use on unit cost and on production labor requirements focuses on the
industries included in the metalworking sector.
What is a metalworking industry? This simply means that all or
most of the establishments classified within the industry are involved
to some degree in the shaping, finishing, and assembling of metal
products.! Which industries are metalworking industries? One way
to answer this question is to identify those industries that use the
"metalworking equipment"-metal-cutting machines, metal-forming
ROBOTICS AND FLEXIBLE MANUFACTURING TECHNOLOGIES 75
machines, joining equipment, and other types of inspection and

finishing equipment. Every five years since 1925, the American
Machinist Magazine has conducted a census of metal-shaping,
metal-forming, and related metalworking equipment. Industries
within the following major SIC groups were included in the
American Machinist Inventory conducted between 1976 and 1978:
SIC code Major group name
25 Furniture and fixtures

33 Primary metals
34 Fabricated metal products
35 Machinery, except electrical machinery
36 Electrical equipment and machinery
37 Transportation equipment
38 Precision instruments
39 Miscellaneous manufacturing
Only the industries in major groups SIC 34-37 are included in

most of the subsequent analysis. These four "core" groups include
over 85% of the units of metalworking machinery counted in the
American Machinist Inventory, and nearly 85% of the total
employment. One major group, SIC 33, primary metals, can be
distinguished from the other major groups because the major activity
of most of its industries is the conversion of unprocessed metal ores
into standard shapes (bar stock, sheets, tubes, pipes, plates, etc.). By
contrast, the primary activity of all the industries in major groups
SIC 34-37 includes either the fabrication, finishing, or assembly of
products from standard metal shapes, and from other purchased parts
and subassemblies. Industries in major SIC 33 are omitted from most
of the subsequent discussion since the metal-refining process is very
different from the processes of fabrication, finishing, and assembling.
Only some of the industries in major groups SIC 25, 38, and 39 are
classified as being in metalworking. Since these industries account for
a relatively small percentage of the machines used and of the people
employed, they too are excluded from the definition of metalworking
and used here.
In 1980, almost 40% of the 20 million people employed in
76 STEVEN M. MILLER
manufacturing and almost 40% of the value added in manufacturing

were concentrated in the four major groups of metalworking
industries, SIC 34-37. About 50% of manufacturing employment
and of value added are concentrated in SIC 33-38.
Vietorisz (1969) has described the metalworking sector as "the
bellwether of economic growth" for an industrial society because ,all
of the tools and capital equipment used by all manufacturing
industries (including itself), and by all other sectors of the economy,
are produced within it. It is the place within the industrial system
where new knowledge is embodied into a physical form, enabling it
to be utilized throughout the entire economic system. Since all new
products and processes require these capital goods, it is not far-
fetched to claim that much of the knowledge that becomes part of the
economic system enters through the metalworking sector. To the
extent that one believes that capital goods, and the role they play in
the creating of new products and processes, are essential to economic
survival and growth, one can argue that the importance of this sector
goes beyond the number of people directly employed within it.
3. THE IMPACTS OF ROBOTIC MANIPULATORS
Surveys of the percentage of workers within selected occupa-

tions that could be replaced by robots have been collected from 22
manufacturing establishments where robots were either being used or
were being seriously considered for use (Table I). These survey
estimates are used to estimate the percent of production workers in
metalworking industries that could be replaced by Level I (insensate)
and by Level II (sensor-based) robots. 2 The survey results are used as
the basis for estimating the percent of jobs in all production worker
occupations that could be performed by robots (Table II).
It is estimated that about 10% of the jobs of manufacturing
production workers could be performed by Level I (insensate) robots
and about 30% by Level II (sensor-based) robots. The Bureau of
Labor Statistics estimates there were 5.1 million production workers
in SIC 34-37 in 1980. Based on the estimates of the percent of jobs
that could be performed by robots and of the number of production
workers, it is estimated that Level I robots could potentially perform

the jobs of 540,000 workers in SIC 34-37 and that Level II robots, if
available, could potentially perform the jobs of 1.45 million workers
in these same industries (Table III). Extrapolating the job dis-
placement data within metalworking to the 14.2 million production
workers in all of manufacturing, it appears that level I robots could
theoretically replace about 1.5 million jobs and that level II robots
could theoretically replace about 4 million jobs. Assuming one robot
replaced two production workers, these estimates of the potential for
robot use, based on an analysis of robot capabilities and job
requirements, imply that there is a potential use for over 700,000
Level I robots or for over 2 million Level II robots throughout all
manufacturing.
Most market forecasts place the cumulative robot population for
1990 within the range of 50,000-150,000 units (Table IV). Assuming
each robot is used to displace two workers, on average, this implies
that only 100,000-300,000 workers will be lost, displacing only
0.7%-2% of manufacturing production workers. Considering only
Level I robots, the estimate of the potential number of applications is
5-15 times larger than the market forecasts of the Level I robot
population for the year 1990.
An attempt is made to explain this large difference between the
estimate of the technical potential for robot use and the estimates of
actual robot sales. The cost of installing robots for loading and
unloading machine tools is analyzed for the purpose of identifying
the conditions under which there would be a strong economic
incentive to use robots given that there is a technical potential for
doing so. The analysis considers the purchase price of the robot as
well as the additional implementation costs that are typically
required. The assumptions made in calculating total implementation
cost, for low-cost, medium-cost, and high-cost robots, shown in
Table V are based on the observations that
• the ratio of total implementation cost/robot base price ranges

from a factor of 3 to 5 in retrofit situations,
• application costs are a larger multiple of the robot base price
for lower-cost robots than for the higher-cost robots,
TABLE I
Summary of Survey Responses of the Percent of Jobs That Could be Robotized by Occupation and by Level of Robot
Technology·
Average Average
Job weighted weighted
ordered by average Number by by
(simple) response for of Minimum Maximum Average distribution batch size
Levell Level responses response response simple of employees distribution
Dip plater 6 20 100 48.3 55.7 43.7

II 6 50 100 78.3 79.7 81.6
Punch press operator I 5 10 100 45.0 44.3 39.0
II 5 60 100 76.0 75.0 67.8
Painter I 16 0 100 40.0 43.5 37.7
II 15 0 100 62.3 66.8 60.5
Rivetor I 3 5 100 38.3 40.5 25.2
I 3 10 100 50.0 51.8 35.9
Shotblaster / sandblaster I 6 10 100 35.8 35.6 31.9
II 6 10 100 35.8 35.6 31.9
Drill press operator 5 25 50 33.0 32.5 30.1
II 5 60 75 67.0 67.0 64.8
Etcher-engraver I 5 0 100 27.0 29.9 24.3
II 5 0 100 53.0 59.2 40.3
Welder I 17 0 60 23.8 25.5 22.0
II 17 10 90 45.6 45.7 47.8
Coil winder 7 0 40 23.6 24.5 24.8
II 7 15 50 38.6 40.2 39.7
Heat treater 3 5 50 21.7 22.7 16.8
II 3 40 90 60.0 61.1 52.5
Machine tool-NC I 20 0 90 19.8 21.7 18.4
II 19 0 100 44.7 46.5 41.2
Grinding/ abrading I 5 10 20 18.0 18.2 19.3
machine operator II 5 30 100 58.0 57.5 53.5
Lathe/ turning I 5 10 20 18.0 18.2 19.3
Conveyor operator I 14 0 50 17.5 14.9 18.7
II 14 15 65 33.2 41.9 33.2
Electroplater I 6 5 40 17.5 18.1 15.2
II 5 15 60 43.0 42.9 44.5
Milling/ planning 5 10 20 16.0 16.1 16.9
Filer/ grinder I buffer I 13 0 35 12.1 9.8 11.6
II 13 5 75 27.7 27.6 26.2
Packager 15 0 40 11.8 10.8 8.7
II 15 0 70 27.1 26.5 20.5
Pourer I 3 5 20 11.7 10.9 13.1
II 3 10 30 20.0 21.4 20.0
Assembler 19 0 40 10.3 8.9 9.5
II 19 15 60 31.1 28.8 29.4
Composites and I 10 10 10.0 10.0 10.0
bonded structures II 40 40 40.0 40.0 40.0
Sheet metal operator 10 10 10.0 10.0 10.0
II 40 40 40.0 40.0 40.0
( Continued)
TABLE I Continued
Average Average
Job weighted weighted
ordered by average Number by by
(simple) response for of Minimum Maximum Average distribution batch size
Level I Level responses response response simple of employees distribution
Inspector I 19 0 25 8.2 7.5 7.9

II 19 5 60 29.2 30.4 28.3
Caster I 4 5 15 7.5 8.6 7.2
II 3 10 20 15.0 15.2 15.9
Electronic wirer 3 0 10 6.7 7.0 7.3
II 3 10 50 30.0 27.6 32.0
Order filler 9 0 20 6.7 6.9 5.5
II 9 0 80 29.4 31.7 25.3
Tester I 17 0 10 5.8 4.8 5.1
II 17 0 30 11.4 11.3 10.8
Mixer I 3 0 10 5.0 6.0 5.2
II 3 10 10 10.0 10.0 10.0
Tender 2 0 10 5.0 6.4 5.5
II 2 20 20 20.0 20.0 20.0
Millwright 4 0 15 3.7 4.4 3.0
II 4 0 15 3.7 4.4 3.0
Kiln-furnace operator 3 0 10 3.3 2.9 2.0
II 3 5 20 13.3 14.7 10.5
Tool and die maker 8 0 5 1.5 1.3 0.9
II 8 0 60 16.6 15.3 9.2
Oiler 3 0 0 0.0 0.0 0.0
II 3 0 0 0.0 0.0 0.0
Rigger 2 0 0 0.0 0.0 0.0
II 2 0 0 0.0 0.0 0.0
Trader/helper 0 0 0.0 0.0 0.0
II 50 50 50.0 50.0 50.0
a For each occupation, the simple average is given by a,imple = (1 In y- I:7~1 p, with n the total number of respondents for the specified occupation and Pi the
ith respondent's estimate of the percent of workers in the given occupation whose jobs could be performed by a robot (Level I or Level II).
For each occupation, the average weighted by size of establishment is given by a,ize = I:;~l sj'"a,imple.j with j the index of size classes of respondents:
j = 1, 1-99 production workers in the establishment; j = 2, 100-499 production workers; j = 3, 400-999 production workers; j = 4, >1000 production
workers. Sj is the percent of metalworking production workers employed in establishments of given size class. For example, s, = 0.216 is the percent of
metalworking production workers employed in establishments with 1-99 production workers. a,imple.} is the simple average of substitution estimates of
respondents who are in size class j.
For each occupation, the average weighted by batch size distribution is given by aba<ch = I:!~l b~a'imple,k with k the index of batch sizes of
respondents: k = 1, custom and small batch production; k = 2, large batch production; k = 3, mass production, b k is the percent of value added in the
metalworking sector produced by industries predominated by batch size class j. For example, b , = 0.554 is the percent of value added in the metalworking
industries produced by industries predominated by small batch production. a,imple, k is the simple average of substitution estimates of respondents who are in
batch size class k.
82 STEVEN M. MILLER
TABLE II
Summary of Suruey Estimates of Potential Displace-
ment of Production Workers: All Occupations·
Percent displacement
Occupation Level I Level II
Tool handlers 27.2 46.7

Metalcutting machine operators 15.5 42.6
Metalforming machine operators 26.2 55.0
Other machine operators 13.2 26.2
Assemblers 8.9 28.8
Laborers 3.8 27.7
Miscellaneous craft workers 2.8 13.2
Maintenance 0.0 0.0
and transport workers
Totals 10.6 28.6
• The average percentage displacement within each group of occupa-

tions is based on an analysis of the occupational employment within
one particular industry group, SIC 351. The percentages would vary
somewhat if they were based on the occupational employment of all
industries with SIC 34-37.
and on the estimates of 1982 robot base prices. The summaries of the
total implementation cost for retrofitting low-cost, medium-cost, and
high-cost Level I robot systems into a factory to load and unload
machine tools is given in Table VI.
3.1. Calculation of Payback Periods Based on Direct Labor Savings

A cost-benefit framework is adopted where the benefits are
narrowly defined as labor savings, and where costs are the total costs
of implementing the robot system. While other benefits are somet-
imes realized when robots are used, such as more consistent and
higher-quality processing, increased throughout, and improved con-
ditions for workers moved out of unpleasant jobs, labor savings are
widely regarded as the primary (and often the only) variable to
consider. Engelberger (1980, p. 103) makes the point quite clearly:
The prime issue in justifying a robot is labor displacement.

Industrials are mildly interested in shielding workers from
TABLE III
Number of Jobs Displaced and Robot Population Implied by Ayres/Miller
Estimates of Potential Displacement
Number of jobs displaced

Potential displacement
Employment
Industries (1980) Level I Level II
Metalworking 5,091,800 539,731 1,456,255

(SIC 34-37)
All manufacturing 14,190,289 1,504,171 4,058,423
(SIC 20-39)
Robot population in metalworking and in all manufacturing

Number of robots assuming 1 robot replaces
Industries 2 Workers 3 Workers 4 Workers
Metalworking, SIC 34-37

Level I 270,000 180,000 135,000
(0.54 million workers displaced)
Level II 728,100 485,400 364,000
All Manufacturing, SIC 20-39
Level I 750,000 500,000 375,000
Level II 2,000,000 1,333,333 1,000,000
hazardous working conditions, but the key motivator is the

saving of labor cost by supplanting a human worker by a robot.
So very much the better if a single robot can operate for more
than one shift and thereby multiply the labor saving potential.
The respondents to the CMU robotics survey (Carnegie-Mellon,

1981) and all other available evidence (Whitney et al. 1981; Robotics
in the U.K., 1981; Ciborra, Migliarese, and Romano, 1980) strongly
supports this view. Respondents to these surveys overwhelmingly
ranked efforts to reduced labor costs as their main motivation for
installing robots. Payback periods are calculated under various
assumptions regarding the total annual cost of a worker and the
number of workers replaced per robot (Table VII). Based on
84 STEVEN M. MILLER
TABLE N
Forecasts of the Population of Robots in the U.S. in
1990·
Cumulative
Source of estimate population
Hunt and Hunt 50,000-100,000

(Upjohn Institute)
Conigliaro 122,000
(Bache, Halsey and Shields)
Aron 94,000-95,000
(Dawia Securities)
University of Michigan/Society of 150,000
manufacturing Engineers Delphi
Survey
Engelberger 150,000
(Unimation, Inc.)
Robot Institute of America 75,000-100,000
a Source: Hunt and Hunt (1982, p. 25).
TABLE V
Assumptions for Calculating Total Implementation Costs
Total implementation cost
Type of robot Base price Robot base price
Lower cost $20,000 4

Medium cost $60,000 3
High cost $100,000 2
TABLE VI
Summary of Cost Assumptions for Retrofitting Level I Robot Systems
Robot Total Operators

hardware Development implementation replaced
cost (R) cost (D) cost (L = R + D) per shift
$20,000 $80,000 $100,000 1

$60,000 $180,000 $240,000 1-2
$100,000 $200,000 $300,000 1-3
TABLE VII
Simple Payback Periods for Level I Robots Based on
Labor Savings
25 K per worker per year

Number of shifts per day
Scenario:
Replacement rate per shift 1 shift 2 shifts 3 shifts
1 robot: 1 worker
25 K robot 4.0 2.0 1.3
60 K robot 9.6 4.8 3.2
100 K robot 12.0 6.0 4.0
1 robot: 2 workers
60 K robot 4.8 2.4 1.6
100 K robot 6.0 3.0 1.9
1 robot: 3 workers
100 K robot 4.0 2.0 1.3
30 K per worker per year

Number of shifts per day
Scenario:
Replacement rate per shift 1 shift 2 shifts 3 shifts
1 robot: 1 worker
25 K robot 3.3 1.7 1.1
60 K robot 8.0 4.0 2.7
100 K robot 10.0 5.0 3.3
1 robot: 2 workers
60 K robot 4.0 2.0 1.3
100 K robot 5.0 2.5 1.7
1 robot: 3 workers
100 K robot 3.3 1.7 1.1
comments in the literature on the economic justification of robots

(Smith and Wilson, 1982), it is assumed that a robot would only be
installed if the projected payback periods were three years or less.
First the cost of installing one robot is considered. This also
includes the case of multiple installations if the cost of installing n
robots is n times the cost of installing one robot. If the three-year
payback period were really a hard and fast rule (which it is not, of
86 STEVEN M. MILLER
course), one would conclude from this simplified analysis that given
the technical feasibility of using Level I robots, the only users would
be
• those plants with enough demand to operate on a three-shift

basis;
• those plants where it was possible to eliminate two or more
workers per shift on two shifts with one robot;
• and those plants that could use the low-cost (low-capability)
robots to eliminate one worker per shift for two shifts.
Taking a conservative outlook, suppose it were the case that one

robot only eliminated one worker per shift, that paybacks were
calculated on a two-shift basis, and that the "heavy duty" robots
were required for most machine loading applications, especially in
"heavy" manufacturing. Payback periods would range between 4 and
5 years, depending on total worker costs. These longer than desirable
payback periods would probably discourage many financial analysts
from giving the "go ahead" on robot application. The conclusion
here is that if one takes the most conservative view of the economics
of robot use (e.g., robots are only viewed as labor savers and must
pay for themselves in a very short time period), then it appears that
too long a payback period (or correspondingly, too Iowa return on
investment) will restrain the growth of robot use over the next
several years. Given the assumptions of this cost-benefit model, the
conclusion is that substantially fewer robots would be installed than
could be used. This would mean that the number of jobs displaced
would be closer to the levels implied by the current market forecasts
than by the survey based estimates of the potential for robot use.
A key assumption in the first cost-benefit analysis is that the
cost of installing n robots is n times the cost of installing one unit.
According to applications engineers and consultants, this is not the
case if additional installations are similar to one another. The
development costs (planning, tooling, design of accessory hardware)
for the second and subsequent applications are lower than for the
first one. If one large establishment were to install many robots, or
correspondingly, if one large firm were to install many robots across
several plants, the average cost per robot would be less than if only
one or several units were installed. In a second series of cost-benefit
calculations carried out, the cost of installing n robots in an
establishment (or across a company) is adjusted so that the develop-
ment cost component of total costs decreases by 10% for each
subsequent installation (Table VIII).
Given this revised cost model, payback periods are calculated by
size of establishment for one industry. Within establishments of a
given size class, the number of workers displaced per establishment is
derived from the total number of workers potentially displaced, the
distribution of employment by size of establishment, and from the
number of establishments within the size class. The result of this
analysis is that payback periods are substantially shorter in the largest
sized establishments (1000 and more production workers) than in the
other size classes (Table IX), the reason being that the average cost
per robot decreases as the number of robots installed increases, and it
is assumed that more workers would be displaced in the establish-
ments with more production workers.
An important conclusion of this analysis is that, given the cost
assumptions, only the largest establishments could justify the use of
robots under the conservative assumptions that one robot replaces a
total of two workers, that "heavy-duty," higher cost, robots are
TABLE VIII
Total and Average Cost for Multiple Installations of Similar Applications
60 K Robota 100 KRobota
Average cost Average cost

Number of Total costb per robot Total costb per robot
robots (x 1000) (X1000) (X 1000) (X1000)
240 240 300 300

9 1642 182.5 2125 236.1
46 4429.6 96.3 6455 140.3
a Base price.
b Total cost of installing n Robots is approximated by In = n"R + D" E~-l (0.9)'. Assume: The
development cost for each successive application decreases by 10% for similar applications. In is
the total cost of installing n robots; Ris the robot base price; D is the development cost for the
first installation.
88 STEVEN M. MILLER
TABLE IX
Payback Periods Based on Production Labor Savings by Size of
Establishment, 25K per Worker per Year
60 K Robot (base price)

Size of
establishment lR:lW:2S· lR:lW:3S lR:2W:2S lR:2W:3S
1-19 9.6 9.6 9.6 9.6

20-99 4.8 4.8 4.3 4.8
99-249 4.5 3.1 3.1 1.6
250-499 4.4 3.1 2.4 1.7
500-999 3.1 2.4 2.0 1.5
1000 and> 1.9 1.5 1.3 1.0
100 K Robot (base price)

Size of
establishment lR:lW:2S lR:lW:3S lR:2W:2S lR:2W:3S
1-19 12.0 12.0 12.0 12.0

20-99 6.0 6.0 6.0 6.0
99-249 5.6 3.9 3.9 2.0
250-499 4.4 4.0 3.1 2.1
500-999 4.2 3.1 2.6 1.6
1000 and > 2.8 2.1 1.7 1.3
a lR: 1W: 2S reads as follows: One robot replaces one worker per shift for two shifts.
required, and that labor savings are the only quantifiable economic
benefit. Smaller-size establishments, with only one or two applica-
tions, would not be able to realize the "scale economies" realized
when multiple units are installed. Payback periods would be too high
to obtain the financial approval for robot use.
Survey responses from 52 members of the Robot Institute of
America indicate that as of 1981, robot use was heavily concentrated
in establishments with 1000 and more production workers. The
survey also showed that these large establishments typically used
many robots. This lends support to the hypothesis that the economic
incentives for robot use are much stronger in the largest-sized
establishments than in the smaller ones. It is also noted that the
automobile industry, which has the largest proportion of production
workers in large establishments, is also the largest user of robots.
(The auto industry also has the highest wage rates of any industry in
SIC 34-37.)
Suppose it were assumed that robot use will continue to be
heavily concentrated in the larger establishment and also in the
metalworking industries (SIC 34-37) until the end of the decade.
Almost 40% of the 5.1 million workers employed in these industries
as of 1980 are located in establishments with 1000 or more
production workers. To displace 10% of these workers would
require almost 100,000 robots, assuming one robot displaces two
people. Clearly, if some robots were used in smaller-sized establish-
ments, as well as outside of the metalworking industries, somewhat
more than 100,000 robots would be required. (If robots were used to
displace 10% of production workers in large establishments in all
manufacturing industries, it would imply a robot population of
180,000 units.) Most market forecasts predict there will be a total of
75,000-150,000 Level I robots in use throughout industry by the
year 1990. It is plausible that these forecasts are predicated on the
assumption that robots will mostly be used within the largest
establishments in the metalworking industries, and that roughly 10%
of the jobs in these establishments could be robotized.
This example shows that there is not necessarily an inconsistency
between the estimate that 10% of production worker jobs could
potentially be performed by Level I robots (implying a potential
market of 750,000 robots) and that there are only expected to be
50,000-150,000 Level I robots in use throughout industry by 1990. It
appears that the key to explaining the difference between the
survey-based estimates of technical potential for robot use and the
market forecasts of the number of robots actually sold is an
understanding of how large a segment of the potential market will
have a strong enough economic incentive to install robots given that
there is a technical potential for doing so.
Will robot use continue to be heavily concentrated in large
establishments in metalworking as it has been, and as many market
forecasters apparently expect that it will be? It is important to know
whether future robot use will follow the same market patterns as past
use to understand the extent of potential labor impacts-whether
10% of a small segment of the work force will be displaced or
90 STEVEN M. MILLER
whether 10% of the total manufacturing work force will be

displaced. An understanding of the likely patterns of robot diffusion
over the next several years would also help to understand whether or
not initiatives might be required to promote robot uses in places
where it might otherwise be indefinitely deferred, such as in
smaller-size establishments.
Is the use of Level II (sensor-based) robot systems going to alter
the extent of robot diffusion and make it necessary to reevaluate the
potential impacts on job displacement? At this point, the answer
appears to be no. Currently, sensor-based systems with enough
capability to acquire randomly oriented parts are substantially more
expensive than Level I systems so the payback periods are much
longer. Sometime within the next several years, the cost of sensor-
based systems will drop substantially, and the answer may be yes. 3
Suppose, as a result of future technological improvements, that the
cost of installing a Level II system was the same as the cost of
installing a Level I system. There are perhaps three times as many
applications for Level II systems as for Level I systems. There would
be more potential applications per establishment, and even the
medium size and smaller establishments would have use for several
(or more) robots. Then installing three times as many robots means
that the average cost per installation would be less than for Level I
robots, assuming, as before, that the total cost of installing a robot
decreases as the number of robots installed increases. Payback
periods would decrease, especially for the medium and smaller size
establishments. If this were to happen, there would be good reasons
for reconsidering the market forecasts that predict that there will be
at most 150,000 robots by the end of 1990. Given the plausibility of
this scenario, future studies of robotic impacts should consider the
rate at which the cost and capabilities of Level II robot systems are
changing.
3.2. The Price Elasticity Argument

An analysis has been made of what would happen to production
worker employment in an industry if robots were used to replace
workers and if a decrease in price resulting from higher levels of
productivity stimulated demand for the industry's output. The

question of interest is whether or not price-induced increases in
demand could be expected to increase total labor requirements by a
sufficient amount to offset job displacement in an industry using
robots. In the first scenario, it is assumed that robots decrease cost
only as a result of reducing labor requirements and that that
throughput is held constant. In the second scenario, it is assumed that
robot use results in a 20% increase in throughput as well as a
reduction in total labor requirements. In this scenario, unit labor
requirements and production costs decrease by substantially more
than in the case where throughput is held constant.
Given assumptions on the decrease in production labor require-
ments, and on the increase in the throughput of the factory, the
amount by which the demand for output would have to increase in
order to reabsorb all workers whose jobs are displaced is calculated.
It is assumed that the demand for output increases as its price
decreases. For every !:1p percent decrease in price, demand for output
is assumed to increase by v percent, where the parameter v is referred
to as the price elasticity of demand. Given the calculated price
change, the magnitude of the price elasticity of demand which would
be required to induce enough of an increase in output so that
employment levels would be maintained is calculated. The magnitude
of this "break-even" price elasticity if of particular interest. Clearly,
if the magnitude of the price elasticity of demand were large enough,
and if there were not limitations on how large the demand for output
could increase, any decrease in labor requirements could be offset by
price-induced increases in the demand for output. The concern
relevant to public policy is whether the calculated values for the price
elasticities of demand required to maintain employment levels are
near the levels of price elasticities normally observed in the "real
world" marketplace.
For both scenarios, the increase in demand required to reabsorb
all displaced workers and the value of the "break even" price
elasticity is calculated with and without assuming that there are job
turnovers as a result of attrition. In the case with attrition, it is
assumed that 15% of the workers in the industry leave the work
force as a result of death, retirement, sickness, disability, etc. during
92 STEVEN M. MILLER
the period in which workers are replaced by robots. When attrition is

considered, a smaller increase in demand and a smaller magnitude of
the break-even price elasticity is required to reabsorb workers
displaced by robots since there are job openings created by job
turnover.
The conclusion on whether or not jobs displaced by robots
could be reabsorbed within the same industry is not conveniently
summarized, since it depends on several variables, including
1. the percent of jobs that are displaced;

2. whether or not robot use increases throughput as well as
decreasing labor requirements;
3. whether or not there is job turnover due to attrition.
First consider the case of no attrition in the work force to focus

on whether price-induced increases in demand, by itself, could be
expected to offset job displacement. If throughput is held constant
and price decreases are due only to decreasing a fraction of
production labor cost (Scenario I, no attrition), it is concluded that
very few of the displaced workers would be reabsorbed. If through-
put were also to increase, thereby causing a larger decrease in price
(Scenario 2, no attrition), it seems that a 10% displacement of jobs
could be offset by price-induced increases in output. However,
demand would have to be relatively price elastic. Without consider-
ing attrition, the conclusion as to whether or not the potential job
displacement of Level I robots could be offset depends on the extent
of the economic benefits of robot use. Price-induced increases in
output would not fully offset the potential job displacement of Level
II robots.
N ow consider the case where the size of the work force
decreases by 15% as a result of attrition over a 3-5 year period. If
throughput is held constant (Scenario 1, attrition), and if 10% or
fewer workers were displaced, job openings from turnovers would
outnumber jobs displaced by robots, even without considering the
effects of price decreases. If the use of robots were to also increase
throughput (Scenario 2, attrition), it appears that a potential dis-
placement of up to 20% could conceivably be offset through the
combined effects of job turnover and price-induced increases in

output. Even with attrition, though, it is unlikely that a potential
displacement of 30% could be offset. The conclusion here is that
Level I robots could be fully utilized in an industry and displace 10%
of the workers over a several year period without resulting in any
unemployment. However, there would still be a significant number
of jobs lost if Level II robots were fully utilized and 30% of the
workers were displaced.
This analysis suggests that a more thorough and precise un-
derstanding is required of how robotics will alter labor requirements
in order to further analyze employment issues in industries using
robots. It is important to know if the economic benefits of robot use
are restricted to savings in labor cost, or whether they might also
increase throughput. It is also very important to know about the rate
of job turnover, since the attrition is often cited as the way of
offsetting displacement effects of robots. More detailed information
on rates of attrition are needed to confirm the conclusion made here
that up to 10% job displacement could be fully offset by attrition if
there is no increase in throughput. This analysis also suggests there is
a need to take a more detailed look at the rate of development of
sensor-based robot systems.
Some of the assumptions underlying these conclusions need to
be clarified in order to make the limitations of this analysis more
transparent to the reader. First, issues relating to changes in skill
requirements for a given occupation or changes in the overall
occupation profile are not considered here. In this simplified
framework, it is assumed that if a production worker is displaced by
a robot and if there is a need for an additional production worker
either as a result of an increase in demand or job turnover, then the
displaced worker can be reabsorbed by the firm. Thus, it is assumed
that the skills required by production workers after the implementa-
tion of robots do not pose a barrier to reabsorbing the displaced
workers.
Second, all aspects of cost changes are not carefully considered.
Increases in capital cost required to install robots is ignored in this
analysis so the calcluated price decreases can only be viewed as upper
bounds. If capital cost was included and the decrease in price was
94 STEVEN M. MILLER
smaller, then the magnitude of the price elasticity required to

maintain employment levels would be larger. If the price elasticities
were larger, it is possible that the percentage of job displacement that
could be offset is smaller than indicated.
Third, only one industry is considered here and interindustry
transactions are ignored. An important characteristic of most metal-
working industries is that they sell most of their output to other
industries (especially to other metalworking industries) to be used as
capital or material inputs. Suppose all industries were to reduce their
cost by a given amount, say 2%, in one period as a result of reducing
labor cost, and bought and sold materials and equipment from one
another. In the next period, all purchased materials and capital
equipment would be 2% cheaper, so each industry would realize an
additional 1%-2% cost reduction. If these interaction effects result in
larger price decreases than are considered there, then the "break-
even" value of the price elasticity required to maintain employment
levels would be smaller than is indicated. If the price elasticity were
smaller, then it is possible that a larger percentage job displacement
could be offset as a result of price-induced increases in demand.
Fourth, relationships between the cost of manufactured goods
and the level of economic activity in other sectors of the economy
which use these goods are not considered. The possibility that
employment losses in manufacturing might be offset by employment
gains in other sectors of the economy which expand as a result of
decreases in the cost of capital and consumer goods is not explored.
More definitive conclusions require that these four factors be
considered.
4. THE IMPACTS OF THE FULLY UTILIZED, FLEXIBLY

AUTOMATED FACTORY FOR BATCH PRODUCTION
The relationship between the level of unit cost and the level of
output produced is examined across 101 different metalworking
industries. An estimate of the pounds of basic metals and of
processed metal inputs purchased is used as a surrogate measure of
the level of output of each industry.4 Value added per unit and units
of output are computed for each industry using pounds of metal
C6 31.6
'Qj
::E
....0
'"
"C
c:
::s
0
•• •
--
Q.
"C
Q)
"C
10.0 •
«
"C
•• •
Q)
::s •
~ • • •
•
3.2
•
•• •
.,.-••.• • •••• ••
•
• '"
1.0
•
...-. .......
,.. ,..
'.. •. •
•
.3
• •• •
•
•
•
... -
•
• •
•••
•
•
•
• •
•
•
.1
•
.0 .1 1.0 10.0 100.0 1000.0

Pounds of Metal/Establishment
FIGURE 1
Value added per pound of metal vs. pounds of metal/establishment for metalworking
industries, 1977.
processed as the standardized unit of output (Figure 1). 5 Regression

relations between unit cost and unit cost components are summarized
in Table X. The basic structural relationships that underlie the shape
of a "neoclassical" long-run unit cost curve for a particular product
are also apparent in the comparison of unit cost versus output across
96 STEVEN M. MILLER
TABLE X
Summary of Regression Results of Unit Cost versus Pounds of
Output/Number of Establishments across Metalworking Industries
Significance Goodness
Output level for of fit
Dependent Constant elasticity output measure:
variable bo b1 elasticity R2 (percent)
Pooled four-digit data set, SIC 34-37: 101 four-digit industries

output = pounds of metal/number of establishments a
kim -0.7243 -0.371 0.99 22.5
( -6.75)b (- 5.49)
lim -0.835 -0.440 0.99 35.5
( -8.95) ( -7.49)
valm 0.401 -0.436 0.99 31.4
(3.96) ( -6.83)
kll -0.056 0.051 0.95 3.1
(-1.41) (2.06)
klemployee 2.160 0.156 0.99 19.9
(44.17) (5.08)
sll -0.473 -0.114 0.99 10.5
(-9.34) (-3.57)
a Unit cost components (dependent variables): kim is the gross value of equipment and
machinery/pounds of metal; !/m is all included production worker costs/pounds of
metal processed; va/m is the value added/pounds of metal; k/! is the gross value of
equipment and machinery/all included production worker costs; k/e is the gross value
of equipment and machinery/total employees; sf! is the salaries/hourly production
worker wages.
Output measure (independent variable): m/e is pounds of metal processed/number
of establishments.
Output elasticity for each cost component is the estimate of b1 in E[In( unit cost
component)] = b o + b1ln(m/e).
b Numbers in parentheses are t ratios for estimate.
industries. These basic relationships are as follows:
1. Capital costs for equipment and machinery per unit of output

decrease across industries as the units of output produced
Increase.
2. Production labor costs per unit of output decrease across
industries as the units of output produced increase.
3. Value added per unit of output decreases across industries as

the units of output produced increase.
4. Machine utilization increases across industries as units of
output produced increase. 6
The implication is that the tradeoffs which most strongly affect the
organization of production within a particular plant-either organiz-
ing to make small volumes of specialized products at a high cost or
organizing to make large volumes of standardized products at a low
cost-are also affecting the organization of production across
industries.
Because it appears that the custom-batch-mass paradigm char-
acterizes the organization of production across industries (as well as
within specific plants), it is argued that the dominant mode of
technology used within an industry can be inferred from the
industry's measures of pounds of metal processed and unit cost.
Industries with the highest levels of production cost per pound of
output and and with the fewest pounds of output are classified as
being comprised of custom and small batch producers. Industries
with the lowest levels of production cost per pound of output and
with the most pounds of output are classified as being comprised of
mass producers. The remaining industries, those with midrange levels
of production cost per pound of output and with midrange levels of
pounds of output, are classified as being comprised of batch
producers. With these assumptions, the proportion of value added
and of output accounted for by metalworking products which are
custom, batch, and mass produced is estimated. The result of this
analysis is that the industries with the highest levels of output and
with the lowest levels of unit cost (which are assumed to be the mass
producers) account for at most 25% of the value added and for less
than 35% of the total output of the 101 industries in the sample 7
(Tables XI and XII). This analysis corroborates the widely cited
claim that most of the value added in the metalworking sector is
accounted for by products which are batch produced.
The significance of the claim that most industries in metalwork-
ing produce batches of specialized products can only be appreciated
by considering the difference in unit cost between batch and mass
98 STEVEN M. MILLER
TABLE XI
Distribution of Value Added by Mode of Production, SIC 34-37
Major SIC groups

Mode of Total,
production 34 35 36 37 34-37
Percent of value added for industries in sample

Custom and small batch 1.1 41.0 58.5 30.6 31.5
Midbatch 28.3 42.5 30.2 6.9 26.4
Large batch 45.5 16.5 3.1 2.9 16.4
Mass 25.1 0.0 8.2 59.6 25.7
Sample coverage of value added in
all industries 94.2 95.8 51.2 97.4 86.0
production. One finding is that a previously published estimate by

Cook (1975) claiming that for the case of a typically machined
product, the unit cost using the most efficient mass production
techniques would be 100-500 times lower than if the produced were
produced in a "one-of-a-kind" mode, and 10-30 times lower than if
it were batch produced seems reasonable. Considering that (1) most
of the value added in metalworking is accounted for by batch
production, and (2) products that are batch produced are much more
expensive than products mass produced in large volumes, it is
typically argued that much of the value added within the metalwork-
ing sector can be viewed as a type of penalty cost that has been
TABLE XII
Distribution of Output by Mode of Production, SIC 34-37
Percent of toal output for

industries in sample
Custom and small batch 1.0 36.3 55.1 20.1 24.8

Midbatch 25.3 44.3 31.4 4.9 23.3
Large batch 45.1 19.4 3.8 2.8 19.9
Mass 28.6 0.0 9.7 72.2 35.7
Sample coverage of output in all
industries 95.1 96.1 51.0 98.0 88.0
unavoidable because of the inherent inefficiencies of custom and

batch production relative to mass production. This is the foundation
for many of the arguments citing the need to accelerate the
development and use of "robotic" and other types of flexible
production technologies which are applicable to batch production.
An analysis is made of the decrease in unit cost and in labor
requirements that would result if the use of robots, in conjunction
with other types of automation, made it possible to substantially
increase the capacity of a factory which uses general-purpose types of
machines to produce specialized products in batches. Previously
published estimates of machine utilization in conventionally or-
ganized factories producing low and mid-volumes of specialized
products (Mayer and Lee, 1980) are used to estimate the potential for
increasing output. The conclusion is that output in batch production
facilities could theoretically be increased by 150%-550% if all of the
productive time available in a year were fully utilized and if the plant
were organized to work more efficiently (Table XIII).
Published information on flexible manufacturing systems in-
dicate that with the most advanced types of flexible automation
currently available, parts of the manufacturing process can be fully
automated even when making specialized products in batches. In
these systems, the output is several times that of its conventional
counterpart, which is consistent with the range of increase derived
from the analysis of theoretical capacity. Robotic manipulators, per
se, are only a very small part of the total automation used in these
plants. This suggests that when analyzing the case of a fully utilized
plant running around the clock, one should more appropriately
address the potential impacts of flexible automation systems on cost
and employment, as opposed to the impacts of robotic manipulators.
Also, these examples indicate that very large capital investments are
required to design and install such systems.
To date, it has not been possible to make a detailed comparison
of the capabilities and economics of flexibly automated plants against
those of conventionally organized ones because the published infor-
mation on the handful of flexibly automated plants throughout the
world is too sparse. Given that the unit cost in the proposed
high-volume batch production plant cannot be directly observed, it
100 STEVEN M. MILLER
TABLE XlII
Summary of Potential Increases in Output
Potential capacity
Increases
Base Robots Robots

Type of plant case only with CAM
High volume
Available hour index· 1.00 1.31 1.31
Throughput inde~ 1.00 1.11 1.39
Output index 1.00 1.45 1.82
Increase in output ('Yo) 45 82
Low-volume: double shift
Available hour indes 1.00 2.17 2.17
Throughput index 1.00 1.16 1.52
Midvolume
Available hour index 1.00 2.98 2.98
Low-volume: single shift
Available hour index 1.00 4.35 4.35
• Available hour index: The relative amount by which the time available for
production could be increased. This includes the effects of recouping the days
per year that the plant is not scheduled for production, as well as recouping the
shifts per day that are idle during those days that the plant is scheduled for
production. One hour per day is alloted for preventive maintenance.
b Throughput index: The relative amount by which the time available for
production could be increased during those times that the plant is operating.
This includes the effects of reducing setup time, loading/unloading time, tool
change time, and idle time.
must be inferred or approximated through some indirect means. The

framework used here for estimating the potential reduction in unit
cost is to assume that a flexibly automated plant producing special-
ized products in small and medium sized batches would have some of
the characteristics of conventional plants producing more standard-
ized products in larger volumes. Thus, the unit cost observed in
industries dominated by plants using conventional (e.g., specialized)
types of automation to make more standardized types of products in
larger volumes is used to infer the level of unit cost in a fully utilized
flexibly automated batch production.
A regression relationship between unit cost and units of output
produced across metalworking industries, estimated in Table X, is
used as a starting point for this analysis. The elasticity of unit cost
with respect to output is used to derive the percent reduction in unit
cost that would result from increasing output. Similarly, the elasticity
of production labor cost with respect to output is used to derive the
percent reduction in production labor requirements that would be
realized with an increase in output.
A more detailed analysis of the variation in unit cost across
industries is also carried out. The explanatory variables used in the
expanded multiple regression model are summarized in Table XIV.
The proposed effect on unit cost is also shown by indicating whether
the sign of the estimated elasticity of unit cost with respect to each
variable should be positive or negative. A key feature of the more
detailed analysis is that two surrogate measures of processing
complexity are constructed which are believed to indicate important
differences in the nature of the processing requirements across
industries. One complexity measure is the average unit cost of the
basic metals purchased by an industry, called the basic metal cost
index (BMCI). An increase in the index means that more expensive
metals are used, which is taken as an indication that the difficulty of
the shaping operations increases. Since more difficult operations
require more capital and/or labor inputs to accomplish, unit cost is
assumed to increase and the proposed sign of this elasticity is
positive. The second complexity measure is the ratio of processed
metal input cost to basic metal input cost. It is argued that this
variable is an indicator of the relative proportions of assembly to
TABLE XN
Summary of Explanatory Variables in Multiple Regression Model a
Effect on
unit cost
(sign of
Factor Variable Notation coefficient)
Pounds of metal processed d In(u)

Level of output mle <0
number of establishments dln(m\e)
Complexity of metal dollars of basic metal d In(u)
shaping activities bmci >0
pounds of basic metal d In(bmci)
= basic metal cost index
dollars of purchased metal d In(u)
Degree of assembly pmbm >0
dollars of basic metal dln(pmbm)
"All included" production worker wages + benefits d In(u)
hourly wage
w ~~->o
production worker hours d In(w)
dollars of metals d In(u)
Material coverage c --<0
dollars of total materials d In (c)
a The regression equation is In(u) = -1.17 - 0.295ln(mle) + 0.983ln(bmci) + 0.488

In(pmbm) + 0.948In(w) - 0.765 In (c)
Standard
deviation T-Ratio =
Variable Coefficient of coefficient coefficient/S.D.
Constant -1.1701 0.4625 -2.53

In(m/e) -0.2947 0.0387 -7.61
In(bmci) 0.9827 0.9827 0.0909 10.81
In(pmbm) 0.4881 0.0490 9.96
In(w) 0.9806 0.1911 5.13
In(e) -0.7644 0.2051 -3.73
The standard deviation of Y about regression line is S = 0.3717 with (101 - 7) = 95 degrees of
freedom; R-squared = 87.0%; R-squared = 86.3%, adjusted for D.F.
metal shaping. It is included to account for the difference between

industries which are primarily involved in shaping and forming
versus those primarily involved in assembly. Two reasons for
believing that a higher ratio of processed metal costs to basic metal
costs indicates a more "complex" process are as follows. First, the
higher the ratio of processed metal inputs to basic metal inputs, the
greater the diversity of material inputs used in an industry. There is

some tendency for the ratio of salary costs/production worker cost
to increase across industries as the ratio of processed metal inputs/
basic metal inputs grows larger. This provides some evidence that it
takes more organizational control and supervision to coordinate
production when there is a larger proportion of processed metal
inputs. Second, the inspection of assembled products requires more
than just the verification of dimensions. Since subcomponents must
be properly integrated with one another, testing is required to verify
that the final product performs its designated functions properly.
With electronics equipment, and computers, this can be a fairly
extensive and complicated process. Also, several other variables
which introduce noise into a cross-sectional comparison, such as
differences in wage rates and in the coverage of material inputs used
to construct the output measure, are introduced into the multiple
linear regression model.
It is observed that industries with low levels of output tend to
use more expensive basic metal inputs than high levels of output.
This suggests that "scaling up" means more than just increasing the
volume of production. There also tends to be a change in the mix of
material inputs as well. It is believed that more standard material
inputs are suggestive of simplified and standardized product designs.
If this were the case, it would indicate that processing requirements
themselves are simplified and standardized as the volume of output
increases. For this reason, it is argued that the unit cost elasticity
derived from the simple regression of unit cost against output
without including the complexity parameters incorporates the effects
of both increasing the average batch size and standardizing the
material inputs. When the BMCI and the other explanatory variables
are included in the regression analysis, the magnitude of the unit cost
elasticity is lessened. The reason for this is that the effects of
increasing the average batch size are separated from the effects of
standardizing the material inputs when the complexity parameters
(principally the BMCI) are included in the multiple regression model.
If it were the case that each product is optimally designed for
ease of manufacture in the flexibly automated batch production
factory, as is typically the case with standardized products made in a
TABLE XV
The Elasticity of Unit Cost Components with Respect in Output
Estimated elasticity unit cost

component with respect to output:
Average
share of With Without
unit cost" complexity complexity
Unit cost component ("!o) variablel variables c
Total value addedd 100.0 -0.295 -0.436

1. Labor value added 55.9 -0.345 -0.461
(a) Production worker costs 35.4 -0.306 -0.440
(b) Salary costs 20.4 -0.406 -0.536
2. Nonlabor value added 44.1 -0.232 -0.408
a Average share of total cost for 101 metalworking industries included in sample.
b Output elasticity for each cost component is the estimate of bl in E[ln(unit cost component») =
be + blln(mle) + b 2 In(bmci) + b 3 ln(1 + pmbm + b.ln(w) + b.ln(c).
C Output elasticity for each unit cost component is the estimate of bl in E[ln(unit cost
component») = b o + blln(mle).
d value added = labor value added + nonlabor value added; labor value added = production
worker costs + salary costs.
conventional mass production factory, one would want the effects

both of increasing batch size and of altering material inputs (and
product design) to be included in the unit cost elasticity. If, however,
each product were not designed to minimize the complexity of
processing requirements, as is typically the case with making
specialized products in a conventional batch production factory, then
one would want to separate the effects of standardizing the material
inputs from increasing the average batch size.
The unit cost elasticities estimated from the regression equations
without and with the complexity parameters are shown in Table XV.
The elasticities of total value added per unit of output are used to
derive high and low estimates of the percent reduction in unit cost
that would result from an increasing output from 50% to 1000% in a
batch production plant (Table XVI). The result is that severalfold
increases in output would lead to a very substantial decrease in unit
production cost. For example, if output were to increase by 100%,
the estimated decrease in unit cost ranges from 18% to 26%. If
output were to increase by 1000%, the estimated decrease in unit cost
ranges from 50% to 65%.
TABLE XVI
Percent Decrease in Unit Cost Derived
from Estimate of Output Elasticity
Percent Percent decrease in unit cost

mcrease assuming elasticity equals:
in output -0.295 4 -0.436 b
50 11.3 16.2
100 18.5 26.1
200 27.7 38.1
300 33.6 45.4
400 37.8 50.4
500 41.0 54.2
1000 50.7 64.8
a Output elasticity derived from estimate of b1 in

E[ln(va/m)] = b o + btln(mle) + b 2 1n(bmci) +
b 3 1n(1 + pmbm) + b 4 1n(v) + bsln(c).
b Output elasticity derived from estimate of b t in
E[ln(va/m)] = b o + btln(mle).
ilunit cost = (I + iloutput)-b 1 - 1 where b t IS
the elasticity of unit cost with respect to output.
It is emphasized that the analysis of the economics of the

flexibly automated factory is more speculative than the analysis of the
economics of robotic manipulators. Since the relationship between
unit cost and the level of output is derived from an analysis of
industries using the current generation of production technology, this
is only an indirect analysis of unit production costs in a factory
making use of the new generation of flexible production technolo-
gies. Hence, the analysis is, at best, suggestive of the economics of
production in a newly designed, flexibly automated plant. It is not
known whether a more direct and detailed analysis would yield the
same conclusions. Nonetheless, if the inference of this analysis is
correct, and it is the case that a flexibly automated batch production
plant would have a substantial cost advantage over a conventionally
organized facility, one would expect that these new types of plants
would rapidly diffuse throughout manufacturing industries.
If unit cost in a fully utilized flexibly automated factory is so
much less than in a conventional factory, one wonders why so few
have been built. Is it too difficult and too expensive to build such a
plant, or is it the result of other less tangible factors? No formal
analysis has been carried out to address this question. However, a
few informal interviews with major manufacturing companies
revealed that several companies have plans on the drawing boards to
build such plants. This suggests that within the next few years, more
attempts will be made to construct flexible manufacturing systems in
the U.S. Some executives commented that organization barriers have
stopped plans for building such plants. One interesting comment is
that there are situations where such a plant would have more capacity
than could be utilized by one division of a company. To be fully
utilized, it would have to be shared across divisions. It has been
suggested that this generates organizational resistance because the
plant is no longer "captive" to one manager.
If a plant could be built that has several times the capacity of a
conventional batch production plant, there is the possibility that
several old plants could be closed down and their production
consolidated into the new facility, which has the flexibility to
produce a mix of different products. This seems to be a likely
scenario if the flexibly automated plant were built in a mature
industry where the potential for market growth was limited. One
example worked out, using the results presented here, shows that if
three plants were closed down and their output consolidated into one
high-volume, flexibly automated plant, total labor requirements
would decrease by 30%-40%. This decrease in labor requirements is
based on the elasticity of unit production labor cost estimated from
the regression analysis (which is based on the use of conventional
types of technologies across low, medium, and high volume in-
dustries). The available information on the existing flexible manufac-
turing systems suggests that the one flexibly automated plant might
have substantially fewer workers than even one of the smaller plants
it replaces. If this were the case, the percentage decrease in total labor
requirements would be much larger.
Since the flexible factory scenario holds the largest promise for
reducing unit cost, and potentially poses the largest threat to
employment in an industry, it warrants more serious analysis.
Further research should focus on a more refined and direct analysis
of the economics of production in flexibly automated factories, and

on forecasts of their use throughout specific industries.
ACKNOWLEDGMENT
This is a condensed version of the Ph.D dissertation, "Potential

Impacts of Robotics on Manufacturing Costs in the Metalworking
Industries," submitted to the Engineering and Public Policy Depart-
ment, Carnegie-Mellon University, in May, 1983. The original
research was supported by the Carnegie-Mellon Robotics Institute,
the Engineering and Public Policy Department, and the Carnegie-
Mellon Program on the Social Impacts of Information and Robotic
Technologies. I am especially grateful to Robert Ayres, committee
chairman, and to Larry Westphal for serving as advisors to the
research which led to the Ph.D dissertation. The present condensed
version was supported by the National Science Foundation, Division
of Policy Research and Analysis, grant No. PRA-8302137 826.
NOTES
1. Throughout this chapter, the Standard Industrial Classification system

(SIC), use by the Bureau of the Census, is used to define industries and
products.
2. For one type of process, metal-cutting machine tool operations, an
estimate is made of the percent of tools that could be operated by Level I
and Level II robots in order to check the validity of the survey estimates.
The two estimates of the potential for robot use in metal-cutting machine
operations, derived independently of one another, are in close agreement.
It appears that the survey-based estimates are good indicators of the
potential for using robots to operate metal-cutting machine tools, and
there is no strong reason to disbelieve the survey-based estimates of
potential robot use in the other application areas either.
3. In a state-of-the-art Level II application for machine loading developed at
the CMU Robotics Institute, much of the added expense was the result
engineering effort required to improve the communication between the
commercially available robot and vision system and the control of the
overall system. The actual Level II hardware, the vision system, only
accounted for a small part of the cost difference. It appears that if the
vendors made minor modifications to their commercially available
systems, it would be possible to achieve the degree of communication and
control required for sophisticated applications without extensive engi-
neering efforts. This would substantially reduce the cost of a Level II
installation.
4. The term "basic metals" refers to inputs of "raw" metal stock-steel,
brass, and aluminum in the form of bars, billets, sheets, strips, plates,
pipe, tubes, etc., as well as casting and forgings made of the three basic
metals. The term "processed metals" refers to inputs which are them-
selves the products of the industries in major groups SIC 34-38. In
general, these products are basic metals which have been further
processed within the metalworking industry.
Pounds of metal processed is divided by the number of establish-
ments within the industry to adjust for differences in the number of
establishments across industries.
There are an additional 31 industries in SIC 34-37 which are
excluded because of inadequate data on their material inputs.
5. Value added/pounds of metal is measured in units of dollars/pound.
Pounds of metal! establishment is measured in units of millions of
pounds/ establishment.
6. This regression result is estimated with data aggregated at the three-digit
SIC level, and is not shown in Table X, which gives results for data
aggregated at the four-digit SIC level.
7. Industries in the sample account for almost 90% of the total value added
in the "universe" of metalworking industries that are considered.
REFERENCES
American Machinist (1978). The 12th American Machinist inventory of

metalworking equipment, 1976-1978, Am. Mach. 122(12), 133-148.
Ayres, R. U., and Miller, S. M. (1983). Robotics: Applications and Social
Implications. Ballinger, Cambridge, Massachusetts.
Bureau of Labor Statistics, U.S. Department of Labor (1982). The National
OES Based Industry-Occupation Matrix for 1980, U.S. Government
Printing Office, Washington, D.C.
Bylinski, G. (1983). The race to the automatic factory, Fortune 107(4),
52-64.
Carnegie-Mellon University (1981). The Impacts of Robotics on the Work-

force and Workplace, Department of Engineering and Public Policy,
Carnegie-Mellon University, Pittsburgh; a student project cosponsored
by the Department of Engineering and Public Policy, the School of
Urban and Public Affairs, and the College of Humanities and Social
Sciences.
Bureau of the Census, Industry Division, U.S. Department of Commerce
(1981). 1977 Census of Manufacturers, MC77-SR-1: General Summary,
U.S. Government Printing Office, Washington, D. C.
(1981). Industry Statistics, Part 3, SIC Major Groups 35-39, Vol. II:
1977 Census of Manufacturers, U.S. Government Printing Office,
Washington, D.C.
(1981). Industry Statistics, Part 2, SIC Major Groups 27-34, Vol. II:
1977 Census of Manufacturers, U.S. Government Printing Office,
Washington, D.C.
Ciborra, c., Migliarese, P., and Romano, P. (1980). Industrial robots in
Europe, Ind. Robot 7(3), 164-167.
Cook, N. H. (1975). Computer-managed parts manufacturing. Sci. Am.
232(2), 22-29.
Engelberger, J. F. (1980). Robotics in Practice, American Management
Association, New York.
Robotics in the UK, Ind. Robot 8(1), 32-38.
Mayer, J. E., and Lee, D. (1980). "Estimated Requirements for Machine
Tools During the 1980-1990 Period," in Machine Tool Systems Man-
agement and Utilization, (Thompson, A. R., ed.), pp. 31-41, Lawrence
Livermore National Laboratory, October, 1980, Vol. 2 of the Machine
Tool Task Force report on the Technology of Machine Tools.
Miller, S. M. (1983). "Potential Impacts of Robotics on Manufacturing Cost
Within Metalworking Industries, Ph.D. dissertation, Carnegie-Mellon
University.
Smith, D. N., and Wilson, R. C. (1982). Industrial Robots: A Delphi Forecast
of Markets and Technology, Society of Manufacturing Engineers,
Dearborn, Michigan.
Vietorisz, T. (1969). UNIDO Monographs on Industrial Development, Vol.
4: Engineering Industry, United Nations Industrial Development
Agency, Vienna.
Whitney, D. E., et al. (1980). Design and Control of Adaptable-
Programmable Assembly Systems, Technical Report R-1406, Charles

Stark Draper Laboratory, Inc.; prepared for the National Science
Foundation, grant No. DAR77-23712.
Yoshikowa, H., Rothmill, K., and Hatvany, J. (1981). Computer-Aided
Manufacturing: An International Comparison, Technical Report, Na-
tional Academy Press; sponsored by the National Research Council.
PART II
INNOVATION AND
MANUFACTURING STRATEGY
CHAPTER 4
INVASION OF A STABLE
BUSINESS BY RADICAL
INNOVATION
James M. Utterback and Linsu Kim
1. INTRODUCTION
How does technological change occur in an organizational,

market and manufacturing context? How does innovation in prod-
ucts and processes affect a firm's competitive viability? Do patterns
of innovation occur from firm to firm or within one firm over time?
If so, what are the implications for research on organizations and for
management? Recent work on innovation in industry contends that
product innovation, process innovation, and organizational structure
and processes become inextricably linked as an organization and its
line of business evolve. The changing relationship among them
stimulates certain directions of further change while it constrains
others or makes them less and less attractive (Abernathy, 1978;
Abernathy and Utterback, 1978). In brief, incremental innovations
albeit with great commercial rewards, become more and more
attractive, while more radical departures, the subject of this chapter,
become diminishingly attractive to established, dominant
competitors.
James M. Utterback • Director of Industrial Liaison, Massachusetts Institute of

Technology, Cambridge, Massachusetts 02139. Linsu Kim. Professor of Manage-
ment, Korea University, Seoul, Republic of Korea.
113
114 JAMES M. UITERBACK AND LINSU KIM
In general terms, the evolutionary model describes the evolution

of products and processes as a transition from an early, fluid state, to
one that is highly specific and rigid. In the early, "fluid" period of
development, performance criteria for new products are not well
defined and market needs or process difficulties are approached
through a variety of different product or equipment designs. Innova-
tion is relatively rapid and fundamental. The production process, in
turn, must be highly flexible, relatively labor intensive, and somewhat
erratIc in work flow (Utterback and Abernathy, 1975).
As development proceeds, however, technological diversity gives
way to standardization. Particular design approaches achieve domin-
ance, production volumes increase, and performance criteria and
processes are more clearly specified. The transition to a "specific"
stage of development entails a change in the nature of innovation. In
contrast to the fundamental changes introduced in the "fluid" phase,
innovation in the "specific" stage is likely to alter only a small aspect
of the basic product, and any changes introduced serve to refine the
established design. On the process side, work flow is rationalized,
integrated, and linear, unlike the fluid and flexible "job shop" of the
early period. Further, general-purpose machines and skilled workers
are replaced by dedicated, highly "specific" equipment (Clark, 1983).
This chapter examines the question of patterns and sources of
radical change as opposed to normal evolutionary change. By radical
we mean change that sweeps away much of a firm's existing
investment in technical skills and knowledge, designs, production
technique, and plant and equipment. Understanding radical innova-
tion requires that we focus on two competing productive units: on
the older technology and on the newer technology.
The present work starts by contrasting the performance and cost
of the old and new products immediately following the invasion of
the new product and process technology. A set of contingent
hypotheses is considered under two major headings-one in which
radical innovation lies with product and product performance, and
the other in which it lies with process and cost. Finally, the
hypotheses developed are further explored through a detailed exami-
nation of several examples.
The authors were stimulated to undertake this analysis by the
INVASION OF A STABLE BUSINESS BY RADICAL INNOVATION 115
observation in their earlier work that evolutionary periods do not

continue indefinitely, but are often interrupted by a new cycle of
creativity and change. Discontinuous changes result in a period of
renewal in a product market segment that results in more vigorous
competition and productivity advance. Such radical changes create
new businesses and transform or destroy existing ones. Substantial
portions of the capital stock will essentially be swept away and
replaced.
An invading technology will have the potential for dramatically
better product performance or dramatically lower production costs,
or both. Figure 1 shows that the performance of a particular product
improves slowly during the period when many alternative design
approaches are being tried. Following the appearance of a dominant
design, however, product performance accelerates rapidly. After
major advances have been made in production process equipment and
integration, a period of more incremental change in product and
process occurs, as indicated by a leveling off of the product
Performance of
Invading Product
Q)
<.J
c: Performance of
§'"
Established Product
.g
Q)
c..
Time
FIGURE 1
Performance of an established product and an invading product contrasted along one
performance dimension. An invading technology will have the potential for dramati-
cally better performance (along some dimensions), although it may be inferior to the
established product at first.
116 JAMES M. UTTERBACK AND LlNSU KIM
performance curve. Continued improvement depends on a new

concept of the product in terms of its design, materials, or principle
of performance.
In reality, however, this figure greatly oversimplifies, because
products must perform along not one, but several or many dimen-
sions. This is indicated in Figure 2 by showing four curves similar to
those in Figure 1, but each with a somewhat different slope and shape
for different performance dimensions. Although some previous
performance limits or constraints may be lifted by the new approach,
others may only be equalled, and perhaps additional constraints
added; thus, the new technology must ordinarily surpass the old on
at least one performance dimension in order to survive. By the same
token, it may perform in some ways that were completely impossible
using the old. Performance improvement in the new technology also
occurs slowly until a better understanding of its form and function
has been gained through trial and experiment in various uses.
New
POl P02
New
P03 P04
FIGURE 2
An invading technology must ordinarily surpass the old in at least one performance
dimension (PD 3 ), but may still be dominated in the others. The new technology may
perform in some ways (PD.) that were completely impossible for the old.
Two critical elements of the argument about discontinuous

change have been implied here. The new technology may be
dominated by the old in many applications for a long period of time.
When it overtakes the old, however, the technical performance of the
new will be accelerating rapidly, while that of the old will only be
improving incrementally. Secondly, the importance of the new
technology may not be as a substitute for the old at all, though it
may ultimately become a substitute, but may lie in new applications
opened up by the dramatic differences between the old and new on
one or more performance dimensions.
Costs follow a similar pattern. Figure 3 shows that unit costs
decline markedly as early gains are made in product design simplifi-
cation, use, and production. Productivity increases continue to occur
at a rapid rate through the transitional period of product
differentiation, development, and major process innovation, with a
Old
!!l
U)
8
...
c:
::J
Time
FIGURE 3
Unit costs of an established product and an invading product over time. The cost of an
invading product, or a product produced by a new production process, may at first be
well above those of an established product or process. At the same time, costs of the
new may fall much more sharply than for the established technology as early gains are
made.
118 JAMES M. UTTERBACK AND LINSU KIM
further decrease in costs. Once a specific pattern has been achieved,

incremental change results in a more moderate decline in costs.
An invading technology will typically have higher costs than the
existing technology, and exhibit a high degree of disorganization and
inefficiency in the production process. At the same time, as early
gains are made, productivity for the invading unit or product will
accelerate much more sharply, and its costs will drop at a cor-
responding rate until they are lower than that of the old.
In general, discontinuous changes focused primarily on product
performance will be quite different in character from those focused
primarily on production cost; the former mainly involve uncertain-
ties outside the firm in terms of the market, use, and performance
advantages of the product, while the latter mainly result in uncertain-
ties within the firm in terms of cost, organization, technical base,
supplier relationships, and so forth. Innovations which result in large
changes in both performance and cost will be the most potent and
disruptive and the most likely to be introduced by productive units
outside an established set of competitors; these types of changes
involve uncertainties both inside and outside an organization and can
create great disarray and resistance.
Following from this argument, four distinctly different patterns
of discontinuous change are sketched out in Figure 4, which shows
the changing frequency of major product and process innovation as
an organization and its technology evolve. If both product and
process are changed, as in the case of functional competition of one
product with another, not only will performance of the old technol-
ogy be surpassed, but cost usually will be as well. If a product alone
is changed, the performance of an existing technology will certainly
be surpassed in some way. If process alone is changed, production
costs will almost certainly be improved dramatically. Finally, if a
process discontinuity results in a great change in product quality or
availability, its performance and usefulness may be greatly increased
as well.
These types of changes will be associated with quite different
patterns in the variables we have discussed. For purposes of outlining
a set of contingent hypotheses, the discussion has been simplified to
consider them under two major headings- one in which emphasis in
PRIMARY CHANGE IN THE PRODUCT
Frequency Frequency
01 Major of Major
Innovation InnovaUon
PRODUCT
-"~OC~
',/.,..............
Time
ENTRY OF NEW ENTRY OF NEW
PRODUCT PRODUCT
PRODUCT·PROCESS PRODUCT DISCONTINUITY
DISCONTINUITY
PRIMARY CHANGE IN THE PRODUCTION PROCESS

Frequency FrequenCy
of Major of Major
_.. / I
"-
I·:';-'''''::~'
" ...... ..... ..... ...
'" .....
Time TI.... I
ENTRY OF NEW ENTRY OF NEW
PROCESS PROCESS
PROCESS DISCONTINUITY PROCESS-PRODUCT DISCONTINUITY
FIGURE 4
Different patterns of discontinuous change on which the contingent hypotheses are
based.
TABLE I
Hypotheses Concerning Discontinuous Change in a Product
Invaded unit will: Invading unit will:
Be an established competitor in the Tend to be a new entrant, either a new

product market segment in question, enterprise or a larger firm carrying its
often producing a standardized prod- technology into a new market area.
uct in large volume.
Generally exhibit the specific or stable Generally exhibit the pattern of fluid or
pattern of characteristics identified in flexible characteristics identified in
earlier research: earlier research. 4
(Continued)
TABLE I Continued
Invaded unit will: Invading unit will:
Strongly perceive the continuing advan- Stress the unique advantages of its
tages of its product in terms of both innovation for some demanding ap-
performance and cost in contrast to the plications or market niches, often not
disadvantages of the invading addressed by the established
innovation. technology.
Tend to view the invading innovation as a Tend to view its innovation as expanding
substitute. the existing market or broadening the
existing range of product uses. It may
later become a substitute.
Often experience a continuing expansion May, by introducing its innovation,
of demand for its product over a actually strengthen demand for the
considerable time (a decade or more). established technology as well for a
time through various complementaries.
Tend to be limited in its responses by Tend to have established strengths in the
large investments in people, equipment, requisite technologies or to acquire
plant, materials, and technical knowl- them via rapid expansion
edge which are irrelevant for success in
the invading technology.
When clearly threatened by the invading Often be in a superior competitive
innovation, redouble its investment in position with respect to skills in the
and efforts to improve its traditional new technology, production and
technology. This will result in a period market understanding, and rates of
of renewed search and creativity and improvement by the time the estab-
often in dramatic improvement of what lished unit begins to respond.
had been a stable technology.
Fail to survive the invastion in most Ultimately dominate the invaded product
cases. However, an established unit market segment. However, it may fail
which has diverse and significant if it does not appropriately evolve
technical skills, and which dominates beyond its initial innovation, a:'d it
its market, may be able to successfully may be acquired by the established
acquire the invading technology. unit, or, in cases, even overwhelmed by
the creative response of the old
technology to the invasion.
Continue to be produced for some In turn, be invaded by yet another wave
specialized market niches for which the of radical innovation.
invading technology is less viable .
• Abernathy and Utterback (1975).

a radical innovation lies with product and product performance, the

other in which emphasis is directed toward process and cost. These
hypotheses are outlined in Table I.
2. A MODEL OF INNOVATIVE CHALLENGES AND

DEFENSIVE RESPONSES
2.1. The New Technology Enters a Market Demanding New

Performance
Radical innovations, like any other type of innovation, require a
recognition of needs and a knowledge of technical and material
resources. Growing dissatisfaction with some aspect of an existing
product, a growing latent demand expressed through an increasing
variety of user experiments with alternatives to existing products or
techniques, or growing public demand through direct product
regulation, may result in discontinuous changes in products. Simi-
larly, clearly evident rewards in terms of market share and profitabil-
ity for dramatically greater productivity, or increasingly tighter
constraints on costs and on resource availability, may result in
discontinuous change in production processes.
In either case, because performance will initially be unreliable
and costs higher, a new technology will tend to start in a relatively
small market where its unique performance advantages are critical-
one ordinarily not occupied or of no great importance to the
producers of the established product. Crude as it is, the new
technology will gain ground by competing in these submarkets, and
its use will expand by means of its capture of a series of them. For
example, mechanical means for making ice were first used in the rural
South where less expensive naturally frozen ice shipped in from the
North was unavailable. As mechanical ice-making technology im-
proved, it became increasingly competitive with natural ice harvested
from rivers and ponds and moved successively northward until the
last commercial operation using harvested ice closed its doors in
Massachusetts in 1948. (Since writing this the authors have been
informed of continuing natural ice harvesting operations today in
Minnesota, Maine, and Canada.)
122 JAMES M. UTTERBACK AND LlNSU KIM
Radical innovations tend to be introduced either by their users,

by small new enterprises entering a business, or by larger firms
diversifying into new markets based on their technological strengths.
We mean by radical innovation one which casts aside the value of
established positions in materials, design, markets, and peoples' skills.
For example, George Eastman was an early amateur photographer
who started his business as a hobby. Because he was not committed
to the ideas and materials used by professional photographers, he
experimented widely with other possibilities for making film. Rayon
was first developed and extruded by a manufacturer of electric
incandescent lamps, the Swan Lamp Company, which needed to
produce more uniform and reliable carbon filaments. One of the first
systems for manufacturing ice was developed by a doctor for medical
uses in his hospital. I
The vacant new markets required for an invasion are sometimes
provided by the performance demands of the federal government.
For example, the transistor (and later the integrated circuit), jet
aircraft engines, and computers all received significant stimulation
from the demands of the Department of Defense, leading to the
formation of new firms to meet the needs for these high-performance
products. Furthermore, federally mandated emission and fuel econ-
omy standards are creating market changes-for example, lighter-
weight automobiles and the use of new materials and electronics in
autos.
Large manufacturers can also act as "external champions" of a
new technology. The demand for new processes for producing large
volumes of plate glass was initially the result of the development of
enclosed automobiles and the Ford Motor Company's consequent
demands for volumes of plate glass. Ford developed the mechanical
elements of a process for continuous casting of glass, but lack of
experience in operating the glass-melting tank caused problems which
Pilkington, an established glass manufacturer in the U.K., helped to
solve. (The relationship of this development, if any, to the later float
glass process is not clear.) Clearly, user needs and sponsorship play
an important role in stimulating radical innovations.
Major corporations can also create reversals in unexpected arenas
by building on their technological skills. For example, the advent of
the diesel electric locomotive was the result of a conscious decision

on the part of General Motors to carry its business beyond its
traditional lines. It heavily backed a number of projects to diversify
around the theme of durable goods connected with motors. Within
these boundaries the diesel venture, in 1930, appeared as an oppor-
tunity to be pursued. The electric typewriter resulted from a similar
urge on the part of IBM to diversify beyond its traditional lines.
2.2. A General Pattern of Response to Radical Innovation

Established productive units often respond to an invasion of
their product line with redoubled creative effort and investment in
what they know well, viewing the new technology as expensive and
relatively crude and believing that it will find only limited applica-
tion. This may be a clear and objective view of the situation at the
time, but it ignores the dynamic potential in the new, as opposed to
the old. The viewpoints of established and invading productive units
are contrasted in Figures 5 and 6. The established technology follows
the performance curve shown at the lower left of Figure 5, first
having won rapid gains, establishing a dominant design, and continu-
ing to improve incrementally. At the same time, the entering new
technology, shown by the right-most curve, can be seen to have
relatively poor performance, which is improving slowly in its early
stages through a disorderly series of experiments and alternative
designs. It is in this situation that an established firm might
emphasize the major difference in performance between the new and
the old along traditional dimensions of measurement, while ignoring
the one or two critical or new dimensions on which the new
technology has a slight or a great advantage, as shown in Figure 2.
The time of entry is shown by a dashed vertical line to facilitate
comparison of the established and contending products.
Turning to Figure 6, the same picture can be seen in terms of
unit cost. The established technology follows the upper and left-most
curve, and at a particular time noted on the abscissa we can see that it
has reached a low and gradually declining level of unit cost and a
relatively high level of quality or performance. At the same time, the
entering new technology shown by the right-most curve can be seen
New
te Time
FIGURE 5
Contrasting the performances of the old and new products along performance
dimension I following the invasion of the new technology. The new technology may
be viewed objectively as relatively crude at first, leading to the belief that it will find
only limited application. This is especially true when it is viewed, often wrongly, as
solely a substitute for the old technology.
to have relatively high unit cost and low quality and performance. It
is in this situation that an established firm might emphasize the major
difference in cost and quality between the new and the old
technology, while the entering firm might emphasize the unique
performance properties of the new technology regardless of cost and
price. As can be seen from Figure 6, this situation may change
quickly as the new technology captures successive market segments
and as production volume expands, and it eventually may dominate
the old technology on the basis of cost as well as performance.
After the new technology bypasses the old, we may see a sharp
break in the old technology as many creative improvements in both
product and process are explored. This is illustrated in Figures 7 and
8. Many elements of the old technology may have been regarded as
fixed following the emergence of the dominant design. Rivalry from
the new technology may lead existing firms to reconsider these
Old
ti
o
u
te Time
FIGURE 6
Contrasting cost of the old and new products immediately following the invasion of
new product or process technology. The new technology may be viewed objectively as
relatively expensive at first. This situation may change quickly as the new technology
captures successive market segments and as production volume expands.
elements and to find that over an ensuing period, possibly of several

decades, dramatic gains have been made in technology, materials, and
scientific understanding that can vastly improve those parts of the
existing product. Established firms may well have organized around
production and improvement of different components or subsystems
of the product. Reconsidering the product as a system again may lead
to much simpler and more elegant designs and rapid gains in
performance and reductions in cost, as variations of each component
or part of the product are considered in terms of their interaction
with the other parts.
Why is it, though, that as threats become clearer with the
improvement in performance and cost of the new technology,
existing productive units so seldom withdraw from the old and
compete aggressively in the new? There are many reasons why new
entrants, as opposed to firms with an existing stake in the business,
should be expected to be major innovators. The rewards may be
greater to the entering firm which views the innovation as opening a
Performance of
Invading Product
Performance of
Established
Q)
u Product
c:
<C
E
~Q)
a..
to
:J
'0
a
c:
Time
FIGURE 7
Creative response in the performance of an established product when surpassed by an
invading product technology. Rivalry from the new technology may result in a
renewal of the performance of the old. The established product may be reconsidered as
a system. Design elements previously fixed may be reexamined. A much broader
search may be entered for improvements.
new market rather than as a substitute for an existing product. The

established firm may view rewards to be obtained from improvement
of the existing technology as more attractive because, given the high
volume of production of existing lines, return on investment in
improvements can be high, rapidly realized, and relatively certain. By
entering the new technology, an existing firm may only be substitut-
ing for its own products, thus reducing the benefits to be gained. In
markets with relatively stable demand, an innovation may even lower
total revenues. Established firms are faced with choosing among
massive investments in new equipment which will result in stable or
declining sales, modifications of current products which can be built
with existing equipment, price cutting or exit from business. But the
new firm without an existing stake sees the innovative product as a
means for market penetration and rapid expansion. Finally, organiza-
tional influences are also major governing factors. The established
Old
'Ui
..
o
(J
'2
:>
Time
FIGURE 8
Acceleration of productivity in the established technology when unit costs of the new
technology become lower. Rivalry from the invading technology may lead to renewed
innovation and remarkable reductions in the unit costs of the established product.
business may be growing moderately or strongly. Peoples' careers in

the organization have been built on a subtle understanding and long
contribution to the development of the established business. Ways to
improve the old may be clearly seen while the potentials in the new
are much more difficult to comprehend either by the established
firms or invading firms. Most importantly, the established firm will
tend to view the new technology simply as a substitute for the old,
and it is objectively a poor substitute at first. Its real potential to
broaden the base of the technology and market may well be hidden
and may develop in completely surprising and unexpected ways as
users experiment with it in various applications and combinations.
Thus, in nearly every case studied, rather than beginning to atrophy
when challenged by a new technology, or adopting that new
technology, the established firms go through a rapid creative surge
and a period of renewed vigor and investment in what is already
known. 2 When confronted with competition from home electric
refrigerators, the makers of iceboxes and ice for home refrigeration
responded immediately. Iceboxes were introduced with far better

insulation and ventilation allowing a given amount of ice to last as
much as three times longer than it did in prior models. Cost of use to
the homeowner was reduced in other ways as well, including
automatic systems for removing melt water from the icebox, systems
for easily cutting cubes from the larger cake of ice, and prescoring of
the hundred pound blocks delivered door to door to assure the
homeowner of a fair and accurate delivery of ice. The iceman's
unshaven appearance was improved, and he was given a uniform and
a canvas bag in which to carry the ice to the kitchen, preventing
puddles on the floor and improving the company's image for reliable
and efficient service. These measures effectively lowered the life cycle
cost of ownership and use of the icebox to the consumer and allowed
the industry to retain a respectable share of the market for a much
longer period than might have been expected (Anderson, 1953; Lund,
1977).
There are many other examples of this type of response by the
old technology. Following the introduction of celluloid roll film by
Eastman Kodak, the Scoville Corporation attempted to penetrate the
same amateur market with small lightweight plate cameras and
developing and printing outfits for amateurs. Perhaps the most
creative response to the invasion of the photographic market by
multiple exposure systems was the development of a magazine
camera in which a pack of rigid plates each moved in turn into the
focal plane as the previously exposed plate was folded into the
bottom of the camera. The makers of steam locomotives for the
railroads introduced two remarkable innovations in response to the
invasion of their traditional markets by the new diesel electric
technology. Both were based on steam turbines, one of which was
mechanically coupled to the driving wheels of the locomotive, the
other of which was coupled to the driving wheels through an electric
generator and electric motor. Five of the turbine-mechanical engines
were actually built, as were three of the turbine-electric, which were
sent around the United States on tour prior to their introduction in
commercial use. All were quietly scrapped a few years later. The
makers of natural fibers such as wool and cotton labored mightily
under competitive stress from synthetic fibers to make their product
more easily used in manufacturers' applications. Their innovations

included ways of grading the fibers to assure uniform staple lengths
in each batch and ways of coating the fibers to make them more
workable and less likely to break when formed into yarn, imitating
properties of synthetic fibers.
Sometimes the creative response of the old technology may be so
vigorous as to drive back or nearly overturn the new technology.
Aluminum engine blocks, which provided better handling character-
istics and fuel economy when used in automobiles, appeared certain
to capture a major part of the market prior to the introduction of
thin-walled cast iron engine blocks by the Ford Motor Company in
response. Thin-wall casting technology allowed a greater transfer of
production equipment and machining and maintenance skills and
soon prevailed over the new aluminum alternative. Prior to the
introduction of the carbon filament electric lamp by Edison, gas
illumination was accomplished by using an open flame like a candle
flame. The disappearance of their business to the newer competitor
stimulated the innovation of the gas mantle, which consisted of a
glowing ceramic structure excited by the flame, to produce much
greater and more pleasing illumination from gas with a much smaller
amount of fuel. The resulting sharp jump in the efficiency and appeal
of gas very nearly bankrupted the fledgling Edison Electric
Company.
The give and take between electric and gas illumination also
illustrates how often the seeds of the new technology or creative
response were sown in a much earlier period but not recognized until
a crisis caused the need to search broadly for alternatives and
solutions. A button of calcium oxide, heated by a gas flame, had
been used to provide the bright light needed for theater illumination
as early as 1820. Our phrase, "to be in the limelight," originates from
this technology, which foreshadowed the idea of the gas mantle.
Similarly, Edison used various metals as experimental filaments in his
lamps, including tungsten, and various gases to serve as envelopes for
filaments, all suggestive of later technologies which were brought to
perfection after the demise of the initial carbon lamp.
Once the old technology begins to modernize itself in response
to the invasion by the new, the emphasis in competition shifts to
product change and away from cost and quality while, at the same
time, prices may drop with extraordinary rapidity, and many new
options and performance dimensions may be available to users. The
total market may even expand as a direct consequence of the invading
innovation. However, this only postpones the inevitable abrupt
decline of the established technology and lends false strength to
arguments against withdrawal from the old and rapid investiment in
the new.
In some product lines, the last few firms in the established
technology can be highly successful and profitable and even highly
innovative. There will probably always be a demand for fine
mechanical watches, and perhaps the few firms that survive the
present shakeout in the industry will be highly profitable and stable
companies. And the few firms which remain manufacturing vacuum
tubes probably supply a highly specialized and profitable market for
high performance designs, research, and other specialized
applications.
All of these statements may be less true in the case of creative
renewals of the production process, because there are fewer un-
certainties in terms of product use and market. Thus, many firms
may appreciate the significance of the gains to be made through
reorganization and reconsideration of the production process itself.
However, one clear difference between product and process
discontinuities is in the source of the radical innovation itself as
outlined in Table II. Earlier, it was stated that users will be more
likely to introduce discontinuous product changes; thus, new firms
and new entrants from other industries'.will usually be the source of
such innovations rather than established firms using the existing
technology. However, established firms will be more likely to
introduce process discontinuities, because they are the creative users
of such equipment. Often, the innovator introducing a radical
process change will be a marginal firm in an industry operating under
severe resource constraints and attempting, by creative means, to
increase its market share and profitability.
The preceding argument has discussed a single wave of change in
performance and cost in a product market segment with a new
technology invading to replace an old. However, as we will see, there
TABLE II
Contingent Hypotheses Contrasting Discontinuities in Product and in Process
Technology
Discontinuous change in product will: Discontinuous change in process will:
Tend to occur in both assembled and Tend to be confined to nonassembled

nonassembled products as defined products.
above.
Usually be introduced by new entrants. Usually be introduced by established
units, often marginal firms seeking to
expand market share or dominant firms
under severe cost, supply, or regula-
tory pressure. May be introduced by
equipment suppliers who may also
enter production in their own right.
Primarily emphasize unique product Primarily emphasize real or potential cost
performance. reduction, improved product quality,
and wider availability.
Require movement toward more flexible Require movement toward more highly
and inefficient or fluid production integrated and continuous production
processes. processes following some characteris-
tics of the transitional pattern described
above.
Often result in smaller scale of operations Often result in larger scale of operations
on introduction. on introduction. Scale may be smaller
in exceptional cases.
Result in rapid growth of demand, Result in rapid growth in demand and
employment, and exports and often, productivity and possibly, though not
though not necessarily, of necessarily, in employment and
productivity. exports.
Result in a high degree of market Not necessarily result in any increase in
uncertainty. The ultimate use of the uncertainty in the market, though uses
product may be quite different from of the product may be broadened.
that initially envisioned.
may be repeated waves of change in product markets-e.g., discrete

transistors replacing vacuum tubes; later, integrated circuits replacing
discrete transistors; and perhaps yet later, microprocessors replacing
specialized integrated circuits. Similarly, there may be waves of cost
reduction as one technology is replaced by another-e.g., harvested
ice being replaced by mechanically produced ice, which in turn was
Frequency Frequency
of Major of Major
Innovation Innovation
Time Time
vacuum tube - transistor discrete transistor - integrated
circuits
piston aircraft engines - turbojets liquid cooled aircraft engine -
air cooled engine turbojet -
turbofan engines
electromechanical calculator mechanical typewriter - electric
- electronic calculator typewriter
ice refrigerators - home mecha nical light open body - closed body
refrigerators automobiles
gelatin glassplate - celluloid roll film rayon - nylon tire cord
steam - diesel electric locomotives air cooled generators - hydrogen
cooled generators
gas lighting - incandescent lamps incandescent - fluorescent lamps
woven carpet - tufted carpet
Frequency Frequency
of Major of Major
Innovation Innovation
/' -- ...... .......
Time Time
open hearth furnace - basic oxygen continuous grinding and polishing

furnace - Pilkington's float glass process
Lubber's cylinder-blowing process natural ice harvesting - mechanical
- mechanical drawing of sheet glass ice -making
primary steel mill - continuous natural gems - synthetic gems
casting industrial gases (oxygen)
coke oven and blast furnace - direct
reduction of iron ores
die drawn wire - drawing from
molten copper
FIGURE 9
Patterns of discontinuous change. Top: Primary change In the product. Bottom:
Primary change in the production process.
replaced by mechanical refrigeration. In the case of mechanical

refrigeration, performance improved as well as cost, whereas when
mechanically produced ice replaced harvested ice, there was no
change whatever in the performance characteristics of the product
but a great change in terms of cost and availability in most of the
market. Waves of entry of new firms, an increasing degree of
competition, a great degree of spinoff of new firms, reduction in
vertical integration, and a wider application of the product or process
will tend to occur with transformations in product and process.
Figure 9 provides some examples of each of the four types of
discontinuities in patterns of innovation outlined in the preceding
argument. We will now explore the ideas expressed above through a
detailed examination of several of these examples, and then ask about
some of the differences in observed patterns based on different types
of threats and different business and technological settings.
3. APPLICATION OF THE MODEL TO SPECIFIC

EXAMPLES
3.1. Product-Process Discontinuities
The pattern of discontinuity created when both product and

process change occurs when small new firms, or larger firms entering
a new business in response to new market opportunities, introduce a
product that is drastically different in its physical or chemical
properties from existing products. Such a new product is often
initially crude and expensive. But crude as it is, it may have great
performance advantages in certain submarkets. As emphasis is placed
on performance improvement, the new product undergoes a
series of frequent, important product changes.
During what we have termed a product-process discontinuity,
the new product is so unrelated in materials and its physical or
chemical properties that the producer can usually take no advantage
of the existing production process or experience gained therefrom.
Thus, the process undergoes a series of frequent and important
changes as a dominant design emerges. In other words, while the
eXlstmg product and its process undergo frequent incremental

changes for improving productivity, the new product goes through a
series of frequent, major product changes, accompanied by similar
changes in process.
There are several cases that appear to suggest the pattern
presented here. The transistor was not an evolutionary extension of
the vacuum tube; rather, it was a revolutionary product introduced
by firms outside the vacuum tube industry, and it subsequently
replaced the latter. When the transistor was discovered, vacuum tubes
were standardized, high volume products. However, they were bulky
and consumed much energy to heat filaments, doubling the problem
in portable products as they required a large-size power source.
Bell Labs, which was not a vacuum tube producer, introduced
transistors into this market with the potential advantage of a
miniaturized size and energy efficiency. But the first transistors
(point contact and alloyed junctions introduced soon after) were so
crude that they operated over a relatively narrow temperature and
limited frequency ranges with high electrical noise. However, crude
as they were, the transistor immediately substituted for vacuum tubes
in hearing aids. A series of frequent major changes immediately
followed that improved the performance and reliability of the
product and increased the productivity of the production process.
These major changes included the grown junction transistor, zone
melting method, diffusion devices, and field effect transistors (Little,
1963).
The silicon grown junction transistor operated at a much higher
frequency, approaching several megacycles, with high temperature
performance. The zone melting technique improved the quality of
raw material so much that it made it possible to manufacture
transistors routinely rather than through painstaking and unreliable
methods. The diffusion techniques developed the first thin-film
transistor, which drastically improved its performance in a higher
frequency range and at a higher power, and opened up a large range
of application. As the diffusion techniques were perfected, it became
possible to produce simultaneously hundreds of transistors by
treating hundreds of units on one slide of material. This reduced
labor costs substantially and laid the seeds of the drastic cost and
price reductions in the late 1950s. The field effect methods virtually
eliminated electrical noise, leading to the first thin-film transistor
with usable properties. Through these major changes in product and
process, the once crude transistor became a longer-lived, more
reliable, smaller, and more efficient product which has displaced
vacuum tubes almost completely (Little, 1963).
The turbojet aircraft engine, introduced by von Ohain and
Whittle, who were outside the aircraft engine industry, was not an
evolutionary extension of the conventional piston engine. Turbogas
engines were developed in an effort to develop a new aircraft engine
that would overcome the technical constraints that prevented the
piston engines from increasing their speed to over 400 mph. Com-
pared to both liquid-cooled and air-cooled piston engines that
became highly refined products through 40 years of evolutionary
development, the first jet engine that flew the Heinkel's He17S in
1939 was very crude. However, it offered a potential to break
through the piston engine's technical constraints; thus, it attracted
the attention of military purchasers. A series of subsequent major
changes, such as axial-flow compressor, two spool arrangements,
variable stators, and new materials, developed the jet engine to a
point where it was ten times as powerful and over five times as
reliable as the piston engine, and subsequently displaced the latter
almost completely in military and civilian markets. 3
The electronic calculator was not an evolutionary extension of
the conventional calculators; it was a revolutionary product intro-
duced by the Sumlock, Wyle, and Sharp Cos., who had no
experience in electromechanical calculator production. Compared to
electromechanical calculators, which had become fairly standard
products produced by highly integrated automatic machines, the first
electronic calculators introduced in 1962 and soon thereafter had high
mortality rates in design and production processes. The product was
so crude that the technical soundness of electronic calculators was
suspected by those who produced electromechanical calculators.
(Majumdar, 1977). However, while electromechanical calculators
underwent few major technological changes in product and process
in the 1960s, electronic calculators underwent a series of frequent,
drastic changes. The Sumlock calculator introduced in 1962 used
vacuum tubes, while the Wyle calculator introduced in the same year
used transistors. The latter used eight circuit boards, each of which
required thousands of separate hand-soldering and wiring connec-
tions performed under a powerful microscope. Even the first
Japanese electronic calculators introduced two years later had as
many as 2600 transistors and diodes, and 2300 resistors and
capacitors, and weighed over 35 lbs. Then, electronic calculators
underwent a series of major product changes from those that had
thousands of discrete semiconductor components, to those that had
tens of integrated circuits, and then to those that had a few MOS/LSI
chips. Accordingly the production processes also underwent a series
of major changes from a long assembly operation of 5000 com-
ponents to a technique similar to LSI production that mounted 39
components on a substrate. The latest method eliminated many
previously necessary assembly steps, thereby realizing greater saving
in the production cost. Through these changes, electronic calculators
became almost trouble free and their average price (of Japanese
models) went down from $1132 in 1965 to $18 in 1975 (Majumdar,
1977). In the process, they have completely displaced expensive and
troublesome electromechanical calculators.
Similar evidence can be found in many other industries. For
example, the home mechanical refrigerator was a revolutionary
product that displaced home ice refrigerators (Anderson, 1953).
Electric razors largely displaced safety razors (Cooper and Schendel,
1976). Dry xerography almost completely displaced wet copying
technology. And celluloid roll film in photography was a revolu-
tionary product that completely displaced gelatin glassplate photo-
graphy and greatly expanded photocopier sales to the amateur market
Qenkins, 1975).
In sum, this pattern of reversal in technological change is
triggered by the introduction of a revolutionary product which uses
different input materials and whose physical or chemical properties
are distinctively different from those of the existing standard prod-
ucts. Such a revolutionary product is most probably introduced, by
someone outside the industry in response to market opportunities of
their novel product ideas. Both the product and its production
processes undergo a series of frequent, major changes in order to
meet new requirements in new market segments and compete with

existing products on the basis of performance as well as price. In
other words, the pattern of technological change in both product and
process shift from the specific state of the existing product to the
fluid state of the new product.
3.2. Product Discontinuities

Product discontinuities occur when a new product, basically
similar to an existing product but with important technological
changes in its critical components, invades markets with a higher
performance in response to new market opportunities. Such a new
product is introduced sometimes by established firms and often by
invaders. Initially expensive compared to the existing product, it also
has potential performance advantages in certain submarkets. Thus, it
undergoes a series of important technological changes in product
similar to the pattern noted in the product process reversal previ-
ously. However, the new product is sufficiently similar to the old one
that manufacturers can, to a large extent, apply their established
techniques and facilities in making the new product. Thus, the rate of
major changes in process is expected to be less than in product and to
tail off rapidly after a series of initial changes are introduced. As the
new product becomes refined and its production process rationalized,
competing on the basis of not only performance but also price, it
rapidly displaces the old product.
Several cases illustrate product discontinuities. A classic example
may be the reversal in the pattern of technological innovation
associated with the change from Ford's Model T to its Model A and
other competitive automobiles in the late 1920s. The period 1900-
1909 saw Ford introduce seven different models of automobiles with
cumulative production volume slightly exceeding 40,000 units. In
1908, the Model T emerged and a year later became the dominant
design in the company's product line. The shift away from product
innovation and the rising importance of major process changes at this
point can clearly be seen in Figure 10. The Model T was produced
for 19 years, with cumulative production volume rising during the
period to exceed ten million units. The consequent reduction in both
EARLY
MODELS MODEL T MODEL A ANNUAL CHANGE
10
PRODUCT INNOVATION
~~
....... -- ...."
,~ , PROCESS
\ INNOVATION
,/' \
,
5
,,
\
\
\
\
I
I \\
I
I
\
1900 1910 1920 1930 1940
FIGURE 10
An example of discontinuous change in products at the Ford Motor Company. [Data
are from Abernathy, W. J. and Wayne, K., (1974), Limits of Learning Curve, Harvard
Business Review, September-October.]
product and process innovation over this period of rising product

standardization, and an implicit shift toward more incremental
cost-reducing types of change, can also be seen. Then, the introduc-
tion of a new technology and functional technological competition
resulted in a period of crisis and discontinuity in the auto industry,
especially for Ford. The advent of ductile steel sheet allowed the deep
drawing of body parts. This, among other factors, enabled the
General Motors Corporation to introduce the enclosed automobile,
providing higher performance at a higher price. Ford soon found its
market share and profit margins dropping as a result of the invasion
of its markets. Falling volume produced heavy penalties because of
the heavy fixed investment by Ford in its efforts to achieve
productivity through plant expansion and vertical integration. Its
attempts to improve its existing product to meet the new competition
further eroded its profitability. The result was that Ford shut down
its operations during 1927, throwing 60,000 persons out of work, and
discarded much of its capital equipment at a cost to the company of
over 200 million dollars in investment in current dollars. The figure
clearly shows that there was an immediate jump in major product
change to nearly the maximum level that had been reached just prior
to 1909. Process innovation can also be seen to have accelerated;

however, the designs were sufficiently similar to the old models that
producers could, to a certain extent, apply their facilities and
experience in making the new models. Accordingly the rate of change
was relatively lower in process than in product. Data available on the
rate of innovation at Ford between 1900 and 1940, as illustrated in
Figure 10, show a pattern similar to one hypothesized in Figure 9.
The air-cooled aircraft engine was a new product that virtually
displaced the liquid-cooled engine in the United States in the late
1920s. The U.S. like other countries, relied completely on the
liquid-cooled engine during World War I and a decade thereafter.
During this period, the liquid-cooled engine evolved from the Wright
brother's 16 hp in-line engine, to the 420 hp Liberty, and then to the
Curtiss D-12. The contemporary liquid-cooled (in effect water-
cooled) engine, however, was heavy and troublesome to put in
operation in winter. In addition, it had many radiator leaks and flaws
in the plumbing, which accounted for 20% of the aircraft troubles at
that time. In the wake of these problems in the liquid-cooled engine,
the air-cooled engine emerged initially in response to the Navy's
need for an engine which was lighter and had a trouble-free cooling
system. With encouragement and funding from the Navy, Lawrance,
a small new firm, introduced an air-cooled engine. It was a small and
experimental model. Soon the venture was absorbed by a larger,
established firm, Wright, which undertook a series of important
product developments in the 1920s. By the end of the decade, the
air-cooled engine had become extremely reliable and virtually re-
placed the liquid-cooled engine in the U.S. (Schlaifer and Heron,
1950). However, the two engines were sufficiently similar in prin-
ciples other than the cooling system that producers could, to a large
extent, apply their equipment and techniques in making the air-
cooled engine. This might be a reason why established Wright was
successful in producing operational engines in volume, once the
product was introduced by the small new firm of Lawrance.
The turbofan aircraft engine was a new product that replaced the
turbojet engine in the 1960s. The turbojet evolved from a subsonic
engine with a simple centrifugal compressor, to one with an axial
compressor, then to an efficient supersonic engine with a two-spool
or variable geometry compressor. The first turbofan engines were
developed by established firms-Pratt and Whitney in the U.S. and

Rolls Royce in Britain. Early models were subsonic, but its improved
efficiency immediately found its application in commerical jetliners.
A series of subsequent changes in the engine, such as the cooled
turbine and high bypass method, resulted in a supersonic turbofan
engine which was far more efficient than the turbojet engines (Taylor,
1970). The rapid development of the engine soon dominated the gas
turbine market. For example, the share of the turbofan in the U.S.
commercial aircraft jet engine market increased from about 15% in
1959 to almost 100% in 1967 (Blackman, 1971). While the turbofan
engine underwent a series of important changes in the late 1950s and
1960s, the production process underwent a lesser degree of change, as
producers took advantage of equipment and experience gained from
the old product in making the new.
Similar evidence can also be found in other industries. For
example, the integrated circuit was a new product that replaced
discrete transistors in many important applications. But its produc-
tion process benefited from techniques developed for transistor
manufacturing (Little, 1963). The electric typewriter was a new
product that subsequently displaced the manual typewriter in the
office and also to a large extent in personal use. But the electrical
typewriter was sufficiently similar to the manual typewriter that the
production process underwent a lesser degree of change relative to
the product (Engler, 1965). Nylon replaced rayon in many applica-
tions including tire cord, but its production benefited from experi-
ence gained from rayon (Yale, 1965).
In short, a product reversal is also triggered by the introduc-
tion of a new product. In this case, the new product makes a
distinctive improvement in performance by incorporating important
changes in critical components, materials, or design of an existing
product. In this sense, the new product is not as radically different
from the old product in the product reversal as it is in the
product-process reversal. Such a new product is sometimes intro-
duced by invaders, but evidence indicates that existing producers are
also an important source of new changes, as in the case of turbofan
engines and new automobile designs. It is often initially crude, but
subsequently undergoes a series of important changes that result in
displacement of the existing product from the market. Since the new
product is substantially similar to the old product, established firms
are in an advantageous position to quickly translate new ideas to
volume production.
3.3. Process Discontinuities

In contrast to the first two patterns of discontinuous change,
process discontinuities usually take place when established producers
bring about major changes in the production process in response to
cost pressures, environmental or safety regulations, or depletion of
raw materials. The focus of major change is simply to meet new
requirements in production process for producing essentially the same
product. A similar pattern is also seen when a major change is
required in the process in order to produce an improved version of
an existing product. In this case, the production process is an
essential part of developing an improved product-as in many
material producing industries-so that, although the objective is to
develop an improved product, the focus of technological change is in
the production process rather than in product performance. This
pattern seems to be seen most often in industries that produce
nonassembled products, such as steel, glass, fibers, chemicals, and so
forth that are used to produce more finished products, or products
somehow "shaped" from such materials, such as glass containers,
nails, etc.
When a new process is developed, either to meet new produc-
tion requirements in producing the same product or to develop an
improved product, such a process is expected to be initially expen-
sive, especially when the new process has little to do with the old
one. A series of major changes take place that improve the process'
efficiency. The product may be improved in quality, but its physical
or chemical properties will remain essentially the same. Materials
used are essentially the same, except when a new process uses new
materials because of the depletion of existing ones. Several cases in
the steel, glass, and similar nonassembled product industries illustrate
process discontinuities.
The basic oxygen furnace is a new steel-making process that
displaced the open hearth furnace, a dominant process for over 50

years. The basic oxygen furnace was introduced in the 1950s, when
the open hearth furnace had evolved to be large and efficient. The
first commercial basic oxygen furnace introduce in 1952 was small
and crude. The process was not competitive with highly refined large
open hearth processes in the region where scrap was available
(Hogan, 1971 )-it was viewed by many large steelmakers as crude,
and it was uncertain whether a scaled-up operation was technically
possible (Dilley and McBridge, 1967). However, a series of major
changes in the process during the 1950s made a large-size operation
possible. The size of the process increased from 35 tons per heat in
1954 to 410 tons per heat by 1972. The basic oxygen furnace and its
cousins, such as the Kaldo converter, the Rotor process, the Ajax
process, and Q-BOF have become far more efficient than the open
hearth furnace and gradually displaced the latter. Today, most new
steel-making facilities are oxygen units.
Similar evidence is also seen in the reversal from the Lubber
cylinder-blowing window glass making machine to mechanical draw-
ing of sheet glass. While the Lubber machine became a highly
efficient cylinder-blowing process, a series of efforts were made to
produce flat sheets of glass by drawing directly from the furnace.
After a series of major changes and improvements, Fourcault,
Colburn, and the Pittsburgh Plate Glass Company commercially
introduced slightly different versions of sheet-drawing machines,
which were far more efficient in production and produced better
quality glass. They subsequently displaced the Lubber cylinder-
blowing machine (Douglas and Frank, 1972).
In the history of steel making, the Bessemer converter was
inadequate in meeting the demands of quality steel. The open hearth
process filled the gap left by the inadequacies of the Bessemer in
many industrial countries. The early open hearth was very crude, but
it grew rapidly through a series of major changes and subsequently
ousted its predecessor. By 1920, much of the present equipment had
been developed. Significant changes included, among other things,
the use of coke-oven gas and tar as fuels, water-cooling systems to
extend furnace life, hot metal mixer, charging machines, and
overhead cranes. By this time, the open hearth accounted for more
than 90% of the total steel production in the U.S.
Continuous casting is a new, cost-saving process that is cur-

rently displacing the primary mill in the steel industry. Developed in
the 1950s, the new technology eliminates several steps in producing
semifinished products-slabs, blooms, and billets-and results in
drastic cost savings through a better yield, labor and energy savings,
etc. In contrast to the primary mills that have undergone little major
change in recent years, continuous casting is currently undergoing
frequent major changes in an effort to develop it as an efficient
process (Tarmann, 1972). Direct reduction of iron ores to sponge
iron is a new technology being researched that is designed to replace
the conventional method of coke oven and blast furnace in an effort
to eliminate the air-polluting coke oven operation and also to reduce
capital outlays in the iron-making process. Its economic viability in
other than oil-producing countries is yet to be seen, but further
development, especially of coal-based methods, might eventually
outdate coke oven and blast furnace technologies in producing iron
(Rose, 1977). In all these cases, the focus of technological change was
to develop a drastically different process in order to meet new
requirements in producing the same or better quality products.
In sum, this pattern of reversal in technological innovation takes
place in the production process. In response to cost pressures,
environmental or safety regulations, or depletion of raw materials,
established producers bring about a radical change in the process of
producing basically the same product in order to meet new require-
ments. In other words, while the product remains in the specific state
of technological innovation, the pattern of technological change in
process shifts from the specific state of the existing process to the
fluid state of the new process. Usually, this pattern of reversal is seen
in industries which produce nonassembled products.
3.4. Process-Product Discontinuities

A few of the cases of radical process change studied appeared to
result in radical changes in the product being produced as well. While
the product is physically the same, it may be so altered in terms of
availability or cost that it finds completely new uses as opposed to
when it is produced in the traditional way. The product and process
pattern of discontinuity is more speculative than the three already
outlined, but certainly deserves mention as a topic for further

research. The pattern is more akin to functional competition and
change, the product-process case mentioned first above, than it is to
changes limited either to the product alone or the production process
alone.
One of the earliest and best examples of this pattern is the
replacement of natural ice harvesting by mechanical ice making.
Machine-produced ice was first introduced into markets where
naturally harvested ice was simply not available in any economic
sense: inland areas of the United States with warm winter climates
which were unable to produce, harvest, and store ice locally. Thus,
ice was essentially a new product; agriculture and food processing
had suffered greatly in the inland South from the lack of any
economic means of refrigeration prior to the development of the
ammonia compression process for ice manufacture. For the first two
decades of its manufacture, from 1870 to 1890, the production of ice
was almost completely for market expansion into areas where the
product had previously been unavailable. The resulting expansion of
refrigerated transportation and centralized food processing to a
year-round basis nationwide resulted in a tremendous expansion in
the demand for both harvested ice and machine-made ice. While it
would be difficult to contend that machine-made ice was different
from harvested ice in any significant physical respect, it was certainly
a revolutionary new product using any other standard of judgment.
The production of synthetic gems presents a very similar case.
While artificial rubies or sapphires are physically and chemically the
same as natural materials in almost every respect, the fact that their
size and purity can be controlled in a relatively low-cost production
process makes them radically different than natural gems produced at
much higher prices. The routine use of sapphires for items such as
watch crystals, or of rubies for laboratory vessels for chemical
reactions at high pressures, depends entirely on their artificial
manufacture. Similarly, the expanded use of industrial diamonds far
beyond that which would have been possible based on natural
supplies alone is also dependent on new processes for producing
rather than mining industrial diamonds. None of these new uses is a
substitution in any real sense of the word, nor, does one suspect,
have natural gems been displaced to any degree in their traditional

uses. Thus, this case also seems to be clearly a discontinuity in the
nature of the product as well as the production process.
A final example of process-product discontinuity is the on-site
production of liquid oxygen for industrial uses replacing the previous
system of centralized production and bulk distribution using tank
trucks of liquid gases. The sale of plants to users, or, in some cases,
the operation of plants on-site for users, as opposed to the
production and distribution of oxygen and nitrogen as primary
products, represented a tremendous change in the industrial gas
business. The development of relatively small-scale plants to
efficiently produce oxygen is clearly an enabling technology in the
development of the basic oxygen furnace for steel making and is
intimately linked to a tremendous increase in the demand for, and use
of, oxygen, with oxygen steel making today representing the largest
segment of the market. In this case, the innovation was brought in by
a completely new firm-Air Products-which was formed for the
purpose of developing the idea. Here, again, the product, oxygen, is
physically the same as the earlier product but the change in process
resulted in completely different availability and economics and use of
the product in ways previously impossible. It is significant in this
case that the military was a strong champion of the idea and the
initial market for small-scale liquid-oxygen plants during the Second
World War, and that support for the Air Products Co. was later
provided and continued by NASA, which represented a majority of
the firm's sales, even into the early 1960s. Utterback and Murray
discuss this form of early support for a radical change by government
purchases more fully in their study on the influence of defense
procurement and sponsorship of research and development on the
development of the civilian electronics industry (Utterback and
Murray, 1977).
As was true of product-process discontinuities, a process-
product discontinuity is disruptive both of the internal workings of a
productive unit and its relationships with its market. In keeping with
the earlier hypothesis about sources of functional product competi-
tion, one might suppose that process-product discontinuities also
may more likely originate from new entrants and established firms
invading a new market; the evidence from the few cases examined
would certainly seem to support this supposition. In contrast, all of
the examples of process discontinuity alone, though some coming
from marginal firms, came from firms within the recognized set of
contending productive units within a particular market segment.
4. SOME CONCLUDING OBSERVATIONS
It might be well advised at this point to mention again that the

hypotheses and generalizations above refer to a productive unit or to
one productive unit contending with another, rather than to firms
consisting of a large number of divisions or productive units or to
industries. Within an industry, one productive unit may be undergo-
ing an invasion and crisis, while others are prospering. Earlier, the
case of the Ford Motor Company under invasion from the all-steel
body car introduced by Dodge and by General Motors was
mentioned. This invasion and response resulted in a period of
declining market share and ultimately discontinuance of business for
retooling for Ford, while at the same time, the share of other
productive units was climbing. Pilkington's participation in the glass
industry expanded dramatically after the development of the float
process, and today 85% of flat glass production worldwide is
licensed under this process. Similarly, with the advent of on-site
production of industrial gases, Air Products' business increased
rapidly, partly at the expense of other firms and partly via expansion
of the market made possible by its technology. It is true that in most
of the cases discussed, however, the losers from the invasion were
grouped together in one set of competitors while the gainers
generally represented a different group of firms being formed or
moving into the market or both. In any event, generalizations about
one industry being invaded by another must be taken with extreme
caution-the reality is that organizations compete with one another,
often within the same industry, with widely varying technologies and
strategies. As a generalization, market shares tend to be stable during
periods of incremental technological change and competition may
also be stable or perhaps declining as measured by the number of
effective firms in an industry. During periods of invasion, market

shares will tend to be less stable and competition greater, as measured
by the number of effective firms in the industry.
It is not necessary that invading and invaded productive units
even be parts of different organizational entities for the hypotheses
above to hold. There are indeed cases in which the invading and
invaded productive unit are parts of the same corporate
organization-the behavior of the respective units is often much the
same as if they were completely independent (however, with the
added complication of additional organizational and political and
power relationships existing within the corporation). The usual case
appears to be that the invading unit within the corporation is quickly
stunted or driven out, though this may well occur for reasons that
appear to be completely objective and rational at the time the
decisions are taken. It may be far easier and more effective from an
organizational and psychological point of view for a firm to create
productive units which take its technological strengths into new
market areas contending with outside competitors rather than
attempting to build a new technological base to replace its existing
products in its traditional markets. If a new productive unit is to
challenge an existing one within a corporation, it usually requires a
high level of sponsorship, protection, and tolerance of the different
requirements for rapid innovation and growth if it is to succeed.
Even if all the productive units in an industry are in a period of
crisis and defensive response to an invading technology, it is quite
possible that a number of them can survive within the newly
established technology. In cases where previously dominant firms
appear to dominate in a new technology, it is those with greater
technical resources, a greater diversity of products related to their
technology, and a significantly large market share which appear to
survive. This was clear, for example, in the case of the Ford Motor
Company, and in the case of General Electric responding to the
invasion of its markets by metal filament lamps from Europe.
There are unusual cases in which a dominant firm creates a
radical innovation replacing its earlier product line in the absence of a
major outside threat: e.g., IBM's innovation of the System 360
computer line replacing earlier models based on discrete components,
RCA's innovation of the color television system replacing its

black-and-white equipment, or Texas Instrument's development of
the integrated circuit. Maidique contends that such exceptional
examples are all cases in which the corporation remains under the
actual, or at least the moral, influence of the original entrepreneur
(Maidique, 1980).
Waves of discontinuous change appear to come more gradually
in some industries and more rapidly in others, as noted earlier. Klein
contends that more rapid regimes of discontinuous change are
associated with industries which repeatedly broaden the definition of
their technology and search for new streams of technology to
reinforce the old and in industries in which entry and competition are
intense (Klein, 1977). These hypotheses would certainly be in strong
agreement with those developed from a more micro and behavioral
perspective here.
Finally, discontinuous change may drastically increase the ag-
gregate demand for the products of an industry. The replacement of
the vacuum tube by the transistor and later the integrated circuit has
increased the sales of the electronics industry from several billions of
dollars to several tens of billions of dollars. The replacement of
piston aircraft engines by turbojets has correspondingly dramatically
reduced the costs and increased the seat miles flown by commercial
aircraft. The advent of the electronic calculator has made such
equipment commonplace rather than something rarely encountered.
The advent of Eastman's Kodak camera and roll system transformed
photography from a small professional market to the large and now
familiar amateur market. Replacement of carbon filament in-
candescent lamps by those based on metal filaments multiplied the
demand for incandescent lamps from twenty million to one hundred
million a year in the United States. The invasion of machine-made ice
tripled the demand for harvested ice from 5 million tons per year to
15 million tons, and adding the volume of machine-made ice would
add several more levels of demand to this. Each revolution in glass
making led to a corresponding sharp increase in aggregate demand
for flat glass, and the advent of on-site production of oxygen led to
more than a doubling in the demand of oxygen and its use.
We have discussed the dynamics of product and process change
within two productive units contending with one another, with some
hints as to how groups of competitors might behave. These issues

should be taken up further and expanded, first using the strong
assumption that all productive units competing in a similar environ-
ment are alike in their technologies and innovative responses and,
then, asking what the implications of that assumption would be for
productive units in the aggregate. While this simplifying assumption
is clearly unrealistic, it may lead to some useful hypotheses which
will roughly correspond with the data on larger groups of firms and
may form a base for a more refined set of contingent hypotheses for
further investigation.
ACKNOWLEDGMENT
The research underlying this paper was supported by the

National Science Foundation, Division of Policy Research and
Analysis under grant No. PRA 76-82054 to the Center for Policy
Alternatives at the Massachusetts Institute for Technology.
NOTES
1. Eric von Hippel's recent work gives other examples and reasons why
users might be expected to play this role (von Hippel, 1978, pp. 30-34).
2. Arnold Cooper and Dan Schendel have studied the strategic responses of
established firms in six industries to functional technological competition.
No threatened firm adopted a strategy of early withdrawal from the old
technology in order to concentrate on the new. Moreover, all but one of
the 22 companies continued to make heavy commitments to the im-
provement of the old technology (Cooper and Schendel, 1976).
3. See Schlaifer and Heron (1950). Information does not seem to be available
on changes in production process for jet engines.
REFERENCES
Abernathy, W. J. (1978). The Productivity Dilemma: Roadblock to Innova-

tion in the Automobile Industry, Johns Hopkins University Press,
Baltimore.
Abernathy, W. J., and Utterback, J. M. (1978). Patterns of industrial

innovation, Tech. Rev. 80(7),40-47.
Anderson, O. E., Jr. (1953). Refrigeration in America: A History of a New
Technology and Its Impact, Princeton University Press, Princeton, New
Jersey.
Blackman, A. W., Jr. (1972). The rate of innovation in the commercial
aircraft jet engine market, Technol. Forecast. Soc. Change, 2, 269-276.
Clark, K. B. (1983). How Industries Evolve: The Interaction of Design
Hierarchies and Market Concepts, Harvard Business School, Working
Paper No. HBS-84-20.
Cooper, A. c., and Schendel, D. (1976). Strategic responses to technological
threats, Bus. Horizons 19, 61-69.
Dilley, D. R., and McBride, D. L. (1967). Oxygen steelmaking-fact vs.
folklore, Iron Steel Eng., 44(10),131-152.
Douglas, R. W., and Frank, S. (1972). A History of Glass Making, Foulis,
Henley-on-Thames, England pp. 154-157.
Engler, G. N. (1965). The Typewriter Industry: The Impact of a Significant
Technological Innovation, Ph.D. dissertation, University of California,
Los Angeles.
Hogan, W. T. (1971). Economic History of the Iron and Steel Industry in the
United States, Vol. 3, Lexington Books, Lexington Massachusetts p. 1549.
Jenkins, R. V. (1975). Images and Enterprise: Technology and the American
Photographic Industry, 1839 to 1925, Johns Hopkins University Press,
Baltimore.
Jewkes, J., Sawers, D., and Stillerman, R. (1969). The Sources of Invention,
2nd ed., MacMillan, London.
Klein, B. H. (1977). Dynamic Economics, Harvard University Press,
Little, A. D. (1963). Patterns and Problems of Technical Innovation in
American Industry: Report to the National Science Foundation,
PB181573, U.S. Department of Commerce, Office of Technical Services;
U.S. Government Printing Office, Washington, D.C., pp. 150-158.
Lund, R. (1977). Making products live longer, Technol. Rev. January,
48-55.
Maidique, M. A. (1980). Entrepreneurs, champions, and technological
innovations, Sloan Management Review Winter, 59-76.
Majumdar, B. A. (1977) Innovations, Product Developments and Technol-
ogy Transfers: An Empirical Study of Dynamic Competitive Advantage,
The Case of Electronic Calculators, Ph.D. dissertation, Case Western
Reserve University, pp. 47-56, 64-65.
Rose, S. (1977). The sponge iron in steel's future, Fortune january, 106-116.
Schlaifer, R. D., and Heron, S. D. (1950). Development of Aircraft Engines
and Fuels, Harvard University Press, Cambridge, Massachusetts.
Tarmann, M. B. (1972). A review-continuous casting development, Iron
Steel Eng. 49(12), 61-67.
Taylor, E. S. (1970). Evolution of the jet engine, Astro Aero 8, 64-72.
Utterback, j. M., and Abernathy, W. J. (1975). A dynamic model of process
and product innovation, Omega 3(6), 639-656.
Utterback, j. M., and Murray, A. E. (1977). The Influence of Defense
Procurement and Sponsorship of Research and Development of the
Civilian Electronics Industry, CPA-77-5.
von Hippel, E. A. (1978). Users as Innovators, Tech. Rev., 80(1), 30-34.
Yale, j. P. (1965). Innovation: The Controlling Factor in the Life Cycle of
the Synthetic Fiber Industry, Ph.D. dissertation, New York University.
CHAPTER 5
MANUFACTURING STRATEGY
Concepts and Methods
Morris A. Cohen and Hau L. Lee
1. INTRODUCTION
The development of strategies directed towards improving the

performance of manufacturing operations has recently become a
critical component of the competitive strategy of many firms.
Productivity improvement, quality enhancement, cost reduction, and
customer service campaigns are all consequences of such strategies.
Today's corporate management has the potential to achieve com-
petitive advantage by adopting innovations such as flexible manufac-
turing process technology, automated material handling equipment,
and computerized production and inventory control systems. Many
firms are also reexamining the sizes, locations, and organizational
structure of their manufacturing resources. These developments are
all included in the set of strategy options associated with the
manufacturing function. Manufacturing strategy is concerned with
the development and implementation of plans which affect the firm's
choice of production resources, the deployment of these resources,
and the design of the infrastructure to control operations activities.
In this chapter we consider concepts and methods associated
with the analysis of alternative manufacturing strategy options. The
Morris A. Cohen. Department of Decision Sciences, The Wharton School, Univer-

sity of Pennsylvania, Philadelphia, Pennsylvania 19104. Hau L. Lee • Department
of Industrial Engineering and Engineering Management, Stanford University, Stan-
ford, California 94305.
153
154 MORRIS A. COHEN AND HAU L. LEE
discussion focuses on the results of a research project which was

concerned with the development of a computerized planning support
system to provide analytical and computational capabilities for the
generation of effective manufacturing strategies. The project was
carried out with the support and participation of a major man-
agement consulting firm (Booz, Allen, and Hamilton), which is
actively involved in the operations management field.
The goals of the project included
1. development of a conceptual framework for manufacturing

strategy analysis,
2. the formulation and initial design of a computerized manu-
facturing planning support system (MPSS), and
3. development and validation of prototype software for mod-
ules to be included in the system.
Activities associated with the first phase of the project focused on a

series of in-depth interviews with senior operations management
consultants who are involved in Booz, Allen, and Hamilton's
manufacturing strategy practice. Their inputs were based on a wide
variety of manufacturing strategy projects. The conceptual
framework which resulted from these efforts relates overall business
strategy to manufacturing strategy. It is particularly concerned with
relating the strategy analysis to the firm's competitive environment.
The system design, developed in the second phase, was directed
towards providing methodological and computational support for the
range of quantitative analyses associated with the framework. The
system design was also based on the collection of software needs,
modeling assumptions, and analytic procedures uncovered in our
review of past manufacturing strategy consulting studies.
Software development was restricted to two modules which are
incorporated in the MPSS framework. These modules were con-
cerned with (1) selecting a configuration design for a firm's network
of manufacturing and distribution facilities, and (2) analysis of the
impact of adoption of new process technologies within a manufactur-
ing plant. Validation procedures included test and field implementa-
tions of the system.
This chapter will introduce the conceptual framework for
MANUFACTURING STRATEGY 155
manufacturing strategy and relate it to recent developments in the

operations literature. We will also discuss the design of the planning
support system and its ability to deal with a set of analytic needs
which are of critical importance in the evaluation of alternative
manufacturing strategies. These needs include (1) analysis of econo-
mies of scale and economies of scope in manufacturing, (2) develop-
ment of procedures which relate the costs of manufacturing to plant
focus, production system flexibility, and product line complexity, (3)
development of models which can link decisions, costs, and perfor-
mance throughout the "supply chain" of manufacturing (raw
material to intermediate product plant to final product plant to
vendor distribution center to customer zone), (4) analysis of the
impact of adoption of new technology, and (5) analysis of strategies
for vertical integration, supplier sourcing, and material managment.
The chapter concludes with a brief discussion of the formulation of
two system modules which have been fully developed and im-
plemented. These modules are named PILOT (Production Inbound
Outbound Transportation model) and PRIME (Process Impact
Evaluation Model). Subsequent papers based on this research project
will describe, in more detail, the design of solution algorithms for the
embedded optimization models in PILOT and PRIME, the design of
the support system software, the computatonal performance of the
algorithms and programs, and the results of a number of system
applications.
The chapter is organized as follows. In the next section we review
manufacturing strategy decisions and performance. Section 3 intro-
duces our conceptual framework for manufacturing strategy. The
fourth section describes the functional specifications of the support
system design. The fifth section briefly reviews the implemented
models software and discusses some preliminary implementation
results.
2. MANUFACTURING STRATEGY: DECISIONS AND

PERFORMANCE
The range of decisions, performance measures, and operations

system linkages associated with the evaluation of a manufacturing
strategy can be quite broad. It is essential to link these decisions with

overall business strategy and with other functional area strategies
under which the firm is operating. A variety of frameworks for the
definition of corporate level, business strategies have been introduced
in the management literature. Porter's (1980) framework includes the
generic strategies of (1) cost leadership, (2) product differentiation,
and (3) market segmentation for each SBU (Standard Business Unit).
Each of these strategies gives rise to different and possibly conflicting
performance requirements on the part of the manufacturing function.
Skinner (1978) has referred to these performance requirements as the
"manufacturing mission." The paradigm developed in Section 3 links
a wide range of manufacturing, planning, and control decisions to the
firm's manufacturing mission and generic business strategy. This
linkage is achieved by specifications of the manufacturing mission in
terms of operationally significant performance measures related to
the following areas:
1. Cost;
2. Quality;
3. Service;
4. Flexibility.
The specific decision areas which are associated with manufac-

turing strategy include the areas of product design, process design,
facility design, control policy specification, and organizational de-
sign. Table I lists a variety of planning decisions and control
functions which can be associated with these areas. In the remainder
of this section we briefly review some recent literature related to each
of these decisions and manufacturing strategy.
2.1. Product Decisions

Product decisions involve definition of attributes for the collec-
tion of goods and services which are to be produced by the firm.
Discovery of those attributes which are in fact demanded in the
market is the responsibility of the firm's market research and new
product development functions (see for example, Sommers, 1979,
TABLE I
Manufacturing Strategy Decisions
Product
• Vertical integration
• Design
• Materials sourcing
• Quality
• Market priorities
• Service
Process
• Production process
• Manufacturing technology and equipment
• Capacities
Facilities
• Configuration design (size and location)
• Distribution system links
Control
• Inventory policy
• Scheduling policy
• Purchasing policy
• Distribution policy
• Quality control policy
• Systems
Organization
• Structure (decentralized/ cen tralized)
• Information (databases, computer resources)
• Incentives
• Performance evaluation
• Human resources
Hopkins, 1977, or Buijs 1979). Translation of product demand

attributes into product manufacturing attributes and the performance
requirements associated with delivering those products is the concern
of the operations function. Thus, for example, designs with demand-
ing performance specifications, or which call for expensive input
materials, could lead to the selection of a production technology and
organizational structure suitable for low-volume, high-cost opera-
tions. Product designs targeted towards the achievement of low costs,
can, on the other hand, lead to the need for high-speed, assembly line
or process-oriented manufacturing systems.
Product decisions that relate to manufacturing include specific

product design features, the degree of vertical integration, material
sourcing strategies, quality goals, and customer service levels. As we
shall see in our discussion of methods for manufacturing strategy
analysis, these product decisions serve to provide constraints and/or
objective functions to the underlying constrained optimization eval-
uation models.
The degree of vertical integration is a particularly important
aspect of product design since it involves complex tradeoffs between
production costs, inbound and outbound transportation costs, qual-
ity control, and manufacturing focus. Specification of which com-
ponents are to be purchased, which are to be fabricated internally,
and how the inputs are to be procured fundamentally affects the
design of the manufacturing system. Examination of the historical
development of American industry and the attendant trends of
consolidation and vertical integration suggests the importance of
these factors for the determination of strategic success.
A fascinating account of these changes is to be found in the book
by Chandler, The Visible Hand (1977), in which the history of U.S.
industry from 1860 to 1914 is described. During this period the
current structure and dominant organizational forms for American
industry, as we know them today were determined. Chandler
suggests that the competitive advantages of economies of speed and of
scale, which were made possible by key technological innovations
(railroad, telegraph, refrigeration, and processing machinery), were
exploited by key individuals who developed large-scale, vertically
integrated management hierarchies.
Williamson's book, Markets and Hierarchies (1975), develops a
behavioral theory of economics which builds on the concept of
transaction costs to explain integration. He suggests that organiza-
tions develop hierarchies to procure inputs and thus avoid the market
when certain structural and environmental factors are in place. These
factors are, in part, affected by the competitive structure of the
supplier's markets and the complexity of the manufacturing task.
Porter's (1980) framework for competitive business strategy also
considers the relationship between a firm and its buyers and sellers.
2.2. Process Decisions
Process decisions are associated with the choice of manufactur-

ing technology. Classes of production processes include continuous
flow, assembly line, flexible systems, and general-purpose job shops.
Management must select the process type appropriate to the manu-
facturing strategy of the firm and the product designs determined
above. For each process type, management must also select the
production equipment and machinery, the physical layout of that
equipment, and the associated material handling systems.
A major contribution which is relevant to process selection and
manufacturing strategy is "The Focused Factory" by Skinner (1974).
Skinner introduces the concept of focus, which involves limiting the
number of products and/or production processes in a plant. The
rationale for focus is that it increases productivity through specializa-
tion. Focus enables a firm to achieve competitive advantage since it
avoids conflicts between disparate missions for different product
groups which are manufactured in the same facility. Skinner also
points out that while the advantages of the focus concept seem clear,
one rarely observes it in actual manufacturing facilities.
Focus requirements form a major link between the firm's
manufacturing mission and the development of manufacturing strate-
gies for the accomplishment of the mission. We will define focus in
our strategy models in terms of (1) the number of process technolo-
gies employed in the plant, (2) the number of products being
produced at the plant, (3) the range of production volumes, or (4) the
range of quality level requirements. Manufacturing system design and
control decisions will clearly be constrained by focus goals. Industry
problems which stem from lack of focus are often observed to arise
because of "product proliferation" which occurs as management
attempts to spread plant overhead over larger volumes and more
products (Schmenner, 1976).
Much of the discussion dealing with process selection and focus
is based on the notions of economies of scale and economies of
scope. Economies of scale are said to exist when it is cheaper for a
firm to produce larger volumes of output. Economies of scope, on
the other hand, occur when it is cheaper for the firm to produce
more products with existing manufacturing facilities. Formally, we
define the firm's product mix by the vector X = (XI' X 2, ... , Xn)
and the total cost function by C(X).
2.2.1. Economies of Scale
If C(AX) ~ AC(X) for any scalar A 2:: 1
then the cost function exhibits economies of scale.

2.2.2. Economies of Scope
If C(XI + X2) ~ C(XI) + C(X2)

for any two output vectors XI, XI 2:: 0,
then the cost function exhibits economies of scope. The implications

of these definitions are illustrated with respect to unit (average) costs
in Figure 1 (see Panzer and Wilig, 1981, for further discussion on
economies of scope).
Scale economies arise because of fixed cost inputs, specialized
equipment, and price breaks for inputs. Scope economies occur when
there exist shareable (joint) inputs that can be allocated over different
products (e.g., supervisors, computer systems, plant equipment).
Manufacturing focus can be defined as the existence of diseconomies
of scope at the plant level.
The particular shape of the firm's cost function is, as noted
above, the key determinant of both scale and scope economies. The
shape of the cost function, is, in turn, affected by production process
and technology choices. In general the cost function will be
nonlinear.
The existence of scope or scale economies suggests that large,
multiproduct plants have an economic advantage over smaller,
single-product plants. As we shall see in our discussion of facility
planning, the empirical evidence suggests that firms do not make
these process and/or facility decisions solely on the basis of the shape
of the production cost function. Methods for selection of plant sizes,
locations, and production processes are a major concern of our
manufacturing strategy evaluation system.
1-
Unit
SCALE ECONOMIES Scale ~ Volume

Capacity
••
•
1-
Unit
SCOPE ECONOMIES Complexity ~ # products

=#technology
=# processes
•
•
FIGURE 1
Cost implications of scale and scope.
In a recent paper, Goldhar and Jellinek (1983) suggest that the

new computer-aided manufacturing technologies (flexible systems,
CAD-CAM, robotics) are advantageous because they possess con-
siderable economies of scope and flexibility. Hence, justification for
process technology investment can now be based on both scale
considerations and the number and diversity of the products.
In spite of the apparent advantages of the new computer-aided
systems, adoption has been slow. Gerwin and Tarondeau (1982)
indicate that in all cases the decisions to adopt flexible, computer-

aided systems were based on anticipated productivity improvements.
Implementation problems arose, however, in the areas of quality
control, accounting (cost analysis), production scheduling, and
maintenance. Thompson and Paris (1982) have also considered the
strategic implications of the new process alternatives. They suggest
that, as a result of the availability of these production process
technological alternatives, scale economies will become less impor-
tant, vertical integration will become more important, manufacturing
resource deployment will move towards smaller, geographically
scattered and focused facilities, manufacturing systems will move
towards process control, more skilled human resources will be
required, and, finally, organizational structures will have to respond
to greater needs for integration. These hypotheses suggest that
process technology innovations must be evaluated in the context of a
comprehensive manufacturing strategy analysis.
Product and process design choices should be reviewed jointly as
each product produced by the firm moves through its individual
market life cycle. Hayes and Wheelwright (1979a, 1979b) recently
introduced a product-process graph which relates these two decision
areas and introduces the concept of a process life cycle. Each axis
describes product and process maturation cycles (see Figure 2). The
selection over time of product designs and process designs is plotted
on the graph. The authors suggest that balanced product/process
development leads to positions on the diagonal.
The product/process graph can be used to illustrate the impact
on product and process designs of manufacturing strategies directed
towards the achievement of focus, flexibility, or exploitation of
economies of scale and scope. The graph indicates how effective
evaluation of any manufacturing strategy requires simultaneous
evaluations of all of the manufacturing strategy related decisions. Our
paradigm builds on this concept and will apply it to the expanded set
of decisions noted in Table 1.
A final point of interest in analyzing process selection is the
issue of manufacturing flexibility. Measures for flexibility include (1)
the mix of products to be manufactured on the equipment, (2) the
alternative routings for these parts, (3) the speed of response to
PRODUCT 1. 2. 3. 4.
LIFE CYCLE
~
:;;.:
LOW VOLUME LOW-MEDIUM IViEDIUM-HIGH HIGH VOLUME,
STAGE c:::
-LOW STANDARDI- VOLUME, VOLUME, -HIGH STANDARDI-
PROCESS ZATION -MANY PRODUCTS -FEW PRODUCTS ZATION, ~
LIFE CYCLE C)
UNIQUE PRODUCTS COMIViODITIES '-l
STAGE
~
~
()
1. JOB SHOP MACHINE SHOP, NONE V)
PRINTER
~
~
()
2. BATCH FLOW CONSTRUCTION "<
(DISCONNECTED) EQUIPMENT
3. ASSEMBLY AUTOMOBI LES,

LINE TELEVISIONS
4. CONTINUOUS NONE OIL, PAPER,

FLOW FOOD PRODUCTS
FIGURE 2
Product-process matrix (Hayes and Wheelwright, 1979).
.....
~
product design changes, and (4) the range of volumes of production

which are feasible and economically attractive. Recently there have
been a number of developments directed towards the evaluation of
flexibility in manufacturing (Stecke et al., 1984). Chatterjee, Cohen,
Maxwell, and Miller (1984) have also developed a general model
framework which can measure the impact of product and process
design decisions on flexibility and on manufacturing cost
performance.
2.3. Facility Decisions

Facility decisions are concerned with specification of the size,
location, and linkages of the productive and logistic resources in
geographic space. The firm must organize these resources into a
network of plants, distribution centers, and warehouses which
produce and deliver products to customers at competitive prices and
with acceptable levels of service and quality. The facility design must
also be responsive to the flexibility and controllability needs deter-
mined by the firm's competitive environment. Dynamic issues of
capacity planning and the allocation of products and processes to
multiple plant locations also arise in this decision area context.
Scherer et al. (1975) reports on an extensive empirical evaluation
of the determinants of plant size and location (based on data drawn
from 12 industries in six countries). The results of this study indicate
that most firms do not make facility or capacity decisions solely on
the basis of scale economies. Rather, these decisions are based on
tradeoffs between production costs and outbound and inbound
transportation costs and capital costs. In addition, characteristics of
the firm's market structure (size, share, concentration) and the
regulatory environment are significant in explaining observed
decisions.
In a more recent study, Schmenner (1982) identified various
multiplant manufacturing strategies that a company can take: Product
Plant Strategy, in which some plants are dedicated to the task of
producing the product for distribution anywhere within the
company's market area; Market Area Plant Strategy, in which the
distribution of each plant's production is confined to a specified
surrounding geographic area; Process Plant Strategy, in which

segments of the full production process are assigned to separate
plants; and the General Purpose Plant Strategy, in which there are
plants which can be assigned to any of a number of responsibilities-
products, market areas, process segments, or a combination. The
paper considers empirical evidence collected from the Fortune 500
companies and investigates the validity of the classification scheme.
The general, multiproduct, multilocation, dynamic decision
model associated with the class of facility and capacity location
decisions is, of course, quite complex. Competitive factors such as
entry deterring behavior and market response are also relevant (see,
for example, Spence, 1977). A suggested normative framework for
this class of decisions involves use of periodic (annual), deterministic,
constrained optimization models to specify structural decisions. The
cash flow and investment return consequences of the structural
decisions can be predicted with the aid of aggregate, dynamic,
stochastic planning models. Similarly, the detailed cost and service
consequences of these decisions can be explored with the aid of
control models for inventory and production scheduling.
2.4. Control Decisions

Control decisions involve selection of policies for the various
mechanisms used to manage production and distribution. Such
policies could, for example, involve the choice between "make to
inventory" vs. "make to order" for finished good scheduling, the use
of "push" vs. "pull" controls for distribution and logistics man-
agement, and selection of software systems for inventory control,
shop floor control, and/or production scheduling. The recent deci-
sion by many firms to adopt Japanese "Just-In-Time" inventory
control procedures is an example of a control policy which will have
profound impacts on the entire operations function. Control policies
also arise in the areas of quality management and work force
management (incentives and worker participation in management).
A recent paper that illustrates the strategic aspects of control is
by Dierdonck and Miller (1980). Manufacturing system performance
is specified by the complexity of the production task, uncertainty,
and toleration for slack. These requirements are related to the

production control system design in terms of (1) information
processing systems capabilities, and (2) degree of system integration
across functional areas and classes of decisions. The authors postu-
lated that there is a causal link between these control system
characteristics and the manufacturing mission. Data were collected
with the aid of a panel of industry experts to examine the validity of
this hypothesis.
In a related article Miller (1981) introduces six categories of
manufacturing control system design decisions which should be
linked to manufacturing strategy: (1) architecture, (2) priorities, (3)
flexibility, (4) focus, (5) technology, and (6) responsibility. Architec-
tural design deals with linkages and integration across control
systems (e.g., purchasing, scheduling, inventory control). Priorities
define the production scheduling sequence dispatch rules (e.g.,
Earliest Due Data, Critical Ratio), and it is suggested that the nature
of such rules should be viewed as a policy decision. Flexibility is
concerned with the response and interactiveness of the actual
(computerized) control system. Focus deals with horizontal integra-
tion of control systems across different product classes. It is
suggested that different product classes may require totally separate
control systems. Technology, in the context of this chapter, is
concerned with production control system attributes (computer
hardware, communication links, and organizational procedures).
Organizational responsibility is concerned with the delegation of
responsibility for system integrity and system control.
The importance of control systems and procedures for the
achievement of manufacturing strategy success is also supported by
the articles by Hall and Vollmann (1978) and Heskett (1977). Hall
and Vollmann discuss the issue of MRP (material requirements
planning) systems, and Heskett is concerned with logistics and
distribution system management. Both papers discuss the idea that
there are controllable operating policies that affect the performance of
these functions in achieving strategic success.
Successful implementation of MRP systems is shown to relate to
strategic factors such as a market position (make to order vs. make to
inventory), product structure, production technology, and working
capital management. Logistics control policies such as postpone vs.

speculate (lead time vs. inventory), service consolidation, and market
differentiation are shown to be an important aspect of strategic
performance goals.
2.5. Organization Decisions

Hayes and Schmenner (1978) describe the fundamental link
between the organizational structure of the firm and its manufactur-
ing strategy. The authors, in particular, define product and process
focus in organizational terms (see Figure 3). A product focused firm
is characterized by different plants producing finished products.
PRODUCT FOCUS
CORPORATE STAFF
PLANT PLANT PLANT

PRODUCT PRODUCT PRODUCT
1 2 3
PRODUCT PRODUCT PRODUCT

MARKET 1 MARKET 2 MARKET 3
PROCESS FOCUS
I CORPORATE STAFF
I
~, ~r ~,
PLANT
PROCESS
...,.. PLANT
PROCESS
...,.. FINISHED
PRODUCT
A B C
FIGURE 3
Product versus process focused organizations (Hayes and Schmenner, 1975).
Organizationally this firm has very different corporate staff respon-

sibilities when compared to a process focused organization in which
plants produce intermediate products. In the product focus case there
is less coordinative responsibility at the central staff level and plants
can operate as profit centers. In the process focus case plants are cost
centers and corporate staff is far more intimately involved in
operational issues such as production coordination and material
logistics.
These organizational issues are important to our concept of
manufacturing strategy analysis. Such analyses ultimately lead to
recommendations which can affect both the degree of product and
process focus. It is critical, therefore, to consider the organizational
design implications of manufacturing policy recommendations. In
particular, consistency between the manufacturing focus and or-
ganizational focus must be maintained. A similar point is made in the
article by Galbraith and Schoemaker (1978). Hayes and Schmenner
(1978) point out the difficulties that can occur when facility structure
and organization structure are mismatched in terms of product/
process focus.
The literature reviewed above, which is related to each of the
manufacturing strategy decision areas, lends support to the working
hypothesis of this project, namely, that development and evaluation
of manufacturing strategies requires comprehensive models for
tradeoff analysis. The complexity of these tradeoffs as indicated by
the above discussion suggests further that a hierarchy of models and
procedures will be required. In the next section we introduce a
framework which addresses these concerns.
3. THE MANUFACTURING STRATEGY PARADIGM
The linkages between the set of decisions described above and

the corporate strategy is illustrated in Figure 4. The paradigm
illustrated in that figure is specifically concerned with the definition
of the manufacturing mission for the firm as a function of both
product and manufacturing resource databases. One can view the
manufacturing mission as a weighted average of performance require-
MANUFACTURING STRATEGY PARADIGM
MANUFACTURING STRATEGY ALTERNATIVES

SEGMENTATION MISSION PARAMETERIZATION EVALUATION
>-
t- if>
UJ
UJ >- :::;
t- i= oa:
:::; iD :::; t-
t- :;
V) " <i x Z
::J UJ U 0
0 ffi ~
if> 0 ~
;1
" "
a1 ,131 Y1 15 1
a2 ,132 Y2 62
~~
EVALUATION DECISIONS
MODEL &
STRUCTURE OUTCOMES
• Visicalc
• IFPS
an I ,I3n Yn
I I bn
.- • Lotus 1, 2, 3
Gaps, Constraints,
Objectives
• Cluster Cost;
Analysis Performance
Analysis
• Regression Analysis
MANUFACTURING
RESOURCE BASELINE
FIGURE 4
A manufacturing strategy paradigm was developed, identifying high potential modeling areas.
ments with respect to various cost, service, quality, and flexibility

measures. Specification of these weights, for each product segment,
serves to define the collection of manufacturing missions required of
the operations function.
Examples of manufacturing missions include the emphasis on
low cost and high quality in consumer electronics and the range from
low-cost, low-performance to high-cost, high-performance strategies
in the automobile industry. The evaluation of alternative manufactur-
ing strategy options focuses on the prediction of performance for the
firm with respect to matters such as cost, profit, and market share for
each product segment. These predictions are based upon specification
of choices for each of the various product, process, facility, control,
and organizational decisions associated with the specific manufactur-
ing strategy options under consideration. Procedures for developing
such predictions form the basis for the model structure incorporated
in the manufacturing planning support system. This model structure
is particularly concerned with the selection of optimal or near
optimal design decisions which will enable the operations function to
meet its manufacturing mission for each product segment.
The evaluation model structure indicated in Figure 4 takes as
inputs (1) manufacturing mission requirements, (2) specifications of
the strategy decision sets, (3) the product market structure, and (4)
the cost/performance capabilities of the manufacturing resource
system. These inputs are based on detailed analyses of the product
and resource data bases. The model structure will be used to specify
manufacturing strategy decisions in a manner that is consistent with
the goals and constraints imposed on the firm by the strategy option
under review, the performance capabilities of existing and proposed
production technologies, and the nature of the competitive market
for the firm's products.
We can illustrate this process by again considering the case of
the consumer electronics industry. As noted above, mission require-
ments include low cost, high quality, and flexibility because of the
rapid pace of technological innovation and short product life cycles.
Manufacturing strategy decisions could involve number and location
of plants, use of offshore manufacturing for partial assembly, process
design (adoption of automated assembly and circuit board processing
equipment), etc. Product design decisions could involve use of

integrated circuits and electronic vs. mechanical switches. Market
structure in industries such as television and radio are relatively
concentrated and include significant foreign competition. The cost!
performance capabilities of each manufacturing site, the location of
distribution centers, and the transportation system links must all be
specified in a manufacturing strategy analysis. As one investigates
these matters, a comprehensive picture of the firm's operation
function emerges. The paradigm can then be used to evaluate specific
proposals affecting plant consolidation, organizational structure,
vertical integration, and new process strategies.
The MPSS design incorporates the evaluation model structure as
well as its required data analysis and support activities. The current
list of functions, models, and procedures identified in the system
design includes the following:
1. Segmentation of the product line into groups appropriate for

manufacturing strategy analysis;
2. Cost!performance analysis to estimate the relationships
between the various manufacturing strategy decisions noted
above and the performance measures incorporated in the
manufacturing mission statement;
3. Production! distribution system design procedures with the
capability to evaluate tradeoffs throughout the supply chain
and which link input flows, manufacturing processes, and
distribution flows;
4. Production process evaluation procedures appropriate for
assessment of the cost, quality, and performance impact of
technological innovations in manufacturing in the context of
existing markets, resources, and control systems;
5. Production! distribution inventory system analysis model to
predict the financial and service level impacts of the manufac-
turing strategy on inventory stock levels throughout the
system; the proposed model will do so in a manner which
captures the complexity of the dynamic, stochastic multi-
echelon system embodied in an integrated manufacturing!
distribution operation;
6. Market response model to relate the performance changes

resulting from the implementation of the manufacturing
strategy, to the competitive environment of the firm.
These evaluation areas can be linked in an MPSS in order to

provide a comprehensive prediction of the overall performance of
alternative manufacturing strategy options (see Figure 5). The link-
ages embodied in this design provide a representation of the firm's
operational function. The approach which we have adopted is based
on a hierarchical model framework which includes a production/
distribution design model, linked to a production/process model and
a multilocation inventory model. These models are then linked to a
market response model which acts as a feedback mechanism for
demand inputs. It is important to note that the models identified
above cover a wide range of decisions, operate in quite different
planning horizons, and deal with the operations function at varying
levels of aggregation. The purpose of the MPSS is to provide a
framework for analyzing both long- and short-term decisions as well
as local and regional impacts of manufacturing strategy.
4. MPSS FUNCTIONS AND DESIGN
In this section we describe the functional specifications of the

various modules to be incorporated in the MPSS. The reader should
note that at its current state of development, prototypes for only two
modules have been developed.
4.1. Segmentation Analysis

Segmentation analysis is concerned with aggregation of the
complete set of products or SKU's into smaller number of segments
for the purpose of manufacturing strategy analysis. Such aggregation
is necessary since the data requirements and computational burden
associated with an exhaustive treatment of each individual product is
prohibitive. It is also the case that consideration of detailed, product
specific information is not necessary for the derivation of valid
strategy conclusions.
OBJECTIVES, CONSTRAINTS, ENVIRONMENT (SCENARIOS)
~
:<:
~
~
I C")
'-l
PRODUCTION/ ~
DISTRIBUTION S2
MODEL CJ
'"
~ ~ REVISED DEMAND
~
;:;j
- - - - - - - -- - -"I
, CJ
PRODUCTION/ ~
INVENTORY
PROCESS SUBMODEL
SUBMODEL MARKET
SHARE,
SALES
I
J~
FACILITIES TECHNOLOGY PRODUCT/PROCESS/ INVENTORY

CONFIGURATION PLANT PLACEMENT
ASSIGNMENTS DECISIONS
,.. MARKET
PERFORMANCE RESPONSE
FLEXIBILITY MODEL
COST SERVICE QUALITY (RESPONSE
TIMES)
COMPETITORS' t
PERFORMANCE ....
~
FIGURE 5
Manufacturing planning support system model structure.
Segmentation or cluster analysis is a classical statistical problem.

The object of a classification scheme is to explain the maximum
amount of observed variance in a set of dependent variables of
interest in terms of intergroup differences. Intragroup differences, on
the other hand, should be minimized. At the same time one would
like to minimize the number of groups that are required. Typically
dependent variables of interest include measures associated with the
four areas of system performance: cost, quality, service, and
flexibility.
Most firms utilize product classification schemes of one type or
another. These schemes typically have been developed for marketing,
accounting, or data processing purposes. They will not, in general, be
appropriate for manufacturing strategy analysis. Possible criteria for
grouping or segmentation in the context of manufacturing include the
following:
1. production technology class;

2. profitability (margin, gross profit);
3. production volume (demand rate);
4. demand variability (uncertainty, trend, seasonality);
5. position in the product life cycle;
6. industry concentration (level of competition);
7. service needs (fill rate, stockout probability, response time);
8. quality needs.
Variables based on these criteria serve to define the independent or

classification variables for the clustering scheme. Segmentation
analysis requires access to product files and customer account
records. It can be carried out informally or with the aid of standard
statistical grouping software. It is important to note that any
statistically derived segmentation scheme requires subjective review
to ensure that managerially inconsistent groups have not been
defined.
4.2. Cost/Performance Analysis

As noted in our literature review, manufacturing productivity is
greatly affected by the relationship between plant size or production
volume and unit manufacturing costs. Economies of scale can arise

from a variety of factors (indivisible inputs, price breaks, local
learning effects, etc.). Complexity of the manufacturing processes,
and the range of products being produced at each plant are also
important in the determination of unit costs. Such economies of
scope can arise from joint inputs (control systems, design depart-
ments, etc.). As size or scope increases it is possible for diseconomies
to arise as well with currently available production technology. As
noted in our review, analysis of the tradeoff between economies of
scale and scope has led some analysts of manufacturing strategy to
promote smaller scale, focused factories.
A major determinant of competive pricing is, of course, the
landed cost of each product to its final customer destination. These
costs are determined by the sum of production costs and those
transportation and facility costs generated throughout the supply
chain. Supply chain costs include the cost of raw material shipments,
intermediate product shipments, plant facility and technology
overheads, finished goods shipments, handling within distribution
center facilities, and inventory holding. As new plants are opened, as
new process technologies are incorporated into these plants, and as
the configurations of distribution and production facilities change in
response to the requirements of the manufacturing strategy, the
drivers of landed cost will also change. Measures of production
performance other than cost will also respond to such changes. The
service impact, will in particular, be affected by decisions concerned
with facility configuration, product design, and choice of tech-
nology.
4.3. Production/Distribution Model

The production/distribution model (PILOT) is concerned speci-
fically with decisions affecting the number of factories, their location,
the number and location of distribution centers, the assignment of
product segments to manufacturing facilities, the management of
inbound and intermediate product flows, the allocation of customers
to distribution centers, and the management of outbound finished
product distribution (see Figure 6).
The model structure is concerned with the complete set of
..
POLICY INPUTS (Constraints)
o FOCUS
l
oUTILIZATION
J
T
• •
DATA INPUTS
oGEOGRAPHICAL STRUCTURE oFACILITIES CONFIGURATION
oCAPACITY CHOICES PRODUCTIONI oPLANT PRODUCTION VOLUMES
oCOST PARAMETERS DISTRIBUTION oMATERIAL FLOWS
oPRODUCT STRUCTURE MODEL oCOSTS (INVENTORY, DISTRIBUTION.
o DEMAND PRODUCTION)
FIGURE 6
PILOT production/distribution model.
facility I capacity choice and control decisions outlined in Table II.

The model is formulated to generate near-optimal decisions with
respect to the minimization of annual costs associated with produc-
tion, purchasing, distribution, and facility overhead (plant and
distribution centers) for network structures of the form indicated in
Figure 7. The optimization problem can be characterized in terms of
TABLE II
PILOT Decisions
• How many plants should there be?

• Where should the plants be located?
• How large should each plant be?
• How much of each finished and intermediate product should be produced at each
plant?
• What are the inbound raw material flows for each raw material from each vendor
source to each plant?
• What are the intermediate product shipments from plant to plant?
• What are the outbound finished product flows from each plant to (1) each
distribution center/warehouse, (2) each customer zone?
• What are the distribution flows between distribution centers/warehouses and
from distribution centers/warehouses to customers?
• How many distribution centers/warehouses should there be?
• Where should distribution centers/warehouses be located?
Raw Material Vendors
Intermediate Product Plants
Final Product Plants
Distribution Centers/Warehouse
Customer Zones
FIGURE 7
PILOT network structure.
a large-scale, mixed integer mathematical program with a nonlinear

objective function. The problem is mixed integer since the variables
include discrete (zero-one) variables representing facility open/close
decisions and customer/DC assignments as well as continuous
production and transportation flow variables. The objective function
is nonlinear due to the assumed relationships between variable
production costs and (1) product mix complexity and (2) the
utilization rate at each plant. The following multiplier model for the
cost function is used in the model:
UNIT PRODUCTION COST(Product, Plant)

= BASE COST(Product, Plant)
X MULTIPLIER(Utilization Rate, # Products)
Since the utilization rate and the number of products are

functions of the entire product mix at the plant, the cost function is
nonlinear.
PILOT is a deterministic, periodic (annual), cost-minimizing
mathematical program. It is based on a class of highly successful
model formulations which have been used in the area of distribution
system planning (see Geoffreon and Graves, 1974). PILOT extends

these model structures since it considers plant open/close decisions,
the attendant production planning decisions, and inbound raw
material and intermediate product flows. The model formulation also
takes explicit account of the scale/scope tradeoffs as noted above.
Because of the size of typical problems (hundreds of cities,
dozens of products, plants, and distribution centers), a heuristic
approach to solving the optimization problem is used. We define the
overall model structure in terms of a hierarchy of sub models in
Cohen and Lee (1984). We also outline the steps carried out in the
current implementation of PILOT to solve this problem structure in
that paper.
Let q = (qt, q2' ... ,qJ) be a configuration of open and closed
plants where q; is 1 if plant 1 is open and 0 otherwise. Then for each
plant configuration q we define ZP(q), ZI(q), and ZO(q) to be the
minimal cost values for the product mix, inbound plus intermediate
product transportation flow, and outbound facility and transporta-
tion solutions respectively. Ij is the annual fixed cost for operating
plant j. The overall problem can be formulated in terms of these cost
functions as follows:
I: PLANT CONFIGURATION MASTER MODEL
min 2: qjIj + ZP(q) + ZI(q) + ZO(q)

j
s.t.
MINP :5 2: qj :5 MAXP
j E ]FIX
where MINP and MAXP are the minimum and maximum number of
plants and ]FIX is the user-supplied set of plants which are to be
fixed open in the analysis.
Since PILOT bases its decisions on an annualized, cost-
minimizing objective, it is clear that additional analysis needs to be
carried out in the complete evaluation of a manufacturing strategy.
At the detailed level, the dynamic, stochastic implications of the
structure and flows generated by PILOT need to be examined in

terms of required inventory investments and customer service
consequences. The yet-to-be-developed production/distribution in-
ventory model will consider such issues. At the other end of the time
spectrum it is necessary to consider the multiyear cost flows, taxes,
and investment returns associated with the facility configuration. If
one uses PILOT as a screening mechanism for the generation of a
small set of alternative designs, then financial planning software
packages can be used for this step of the analysis. Alternatively one
could reformulate PILOT as a dynamic optimization problem.
Because of the size and complexity of the resulting class of problems
the former approach was adopted in the MPSS design.
4.4. Production Process Model

The production process model (PRIME) reduces the manufac-
turing process within a factory into a series of stages. (See Figure 8.)
Each of these stages may each have alternative process technology
capabilities which are either in place or under consideration. The
purpose of PRIME is to predict the impact of the alternative process
technology choices on both product unit costs and related measures
of manufacturing system performance (i.e., utilization, product focus,
and process focus). PRIME incorporates an optimization model with
the capability to account for the impact of scale and scope economies
at the process center level in a menu-driven, user friendly system
framework (see Figure 9). The optimization model in PRIME
allocates product volumes to alternative process technologies at each
stage in the manufacturing process. As in the case of PILOT and
PRIME, the inclusion of this provision via the multiplier model
makes the underlying mathematical program nonlinear.
Let p, t, and k represent indices for the manufacturing process
(stage), technology alternative and product class, respectively. Dis-
crete choice variables qpt and qptk are defined as follows:
if technology t is available for process p

otherwise
......
0()
Cl
Process 1 Process 2 Process 3
~r~~u:.t.!._~ Technology 1 ------~ Technology 1 I Technology 1
• Product k ~: :
. . I· : 1-----.
Technology n1 Technology n2 Technology n3
~
Process 4 o~
~
Technology 1 ~
•• ::..
• ()
Technology n4
FIGURE 8
~::..
PRIME model structure. ~
5:c:::
t--
t--
t'>j
t'>j
..
POLICY INPUTS (COST FUNCTIONS)
eEXPERIENCE
eFOCUS
eUTlLIZATION
L )
l"
DATA INPUTS OUTPUTS

ePLANT PRODUCTION ePROCESS SELECTION
VOLUME ePRODUCT/PROCESS
ePROCESS ALTERNATIVES PRODUCTION/ PRODUCTION
eCOST (VARIABLE, SEMI· PROCESS e PRODUCTION COSTS
VARIABLE, FIXED) SUBMODEL ePERFORMANCE
ePERFORMANCE FUNCTIONS
(VOLUME, FOCUS)
(ONE PLANT)
FIGURE 9
PRIME production/process model.
and
if product k is allocated to
qptk = {01 technology t within process p
otherwise
Let \j,(q) be the minimum variable production cost for process p

given the values for qpt and qptk specified in vector q. The overall
choice problem is
s.t.
2: qptk ::5 PFLpt

k
qpn qptk are (0, 1)
where F is the fixed cost of technology p, process t and TFL and PFL
are the upper limits on the number of technologies per process

(technology focus limit) and the number of products assigned to
process p, technology t (product focus limit), respectively. The
process production problem is decomposable for each p. We define
V(q) as
Yp (q) = min 2: Yptkx;,tk

t,k
s.t.
2: rptkx;,tk ::; CAPptqpt

k
2: x;,tk ;:::: Rpk

t
Rpk = Ppk~
Jfptkqptk ::; x;,tk ::; ~tkqptk

x"tk ;:::: 0
[qptk are (0, 1) and fixed]
where Rpk are production requirements for product k at process stage

p; x;,tk is the production volume for process p, technology t, and
product k; Yptk is the variable production cost for ptk; %ptk> ~tk are
the lower and upper bounds on production volume; rptk is the
capacity utilization rate for product k at Pt; Ppk is the routing
indicator variable (Ppk = 1 if product k is processed at process p and 0
otherwise); ~ are production requirements (plant output) for
product k; CAPpt is the capacity for technology t, process p.
4.5. Market Response Model

The collection of objectives, constraints, and environmental
factors which form the inputs to the MPSS model framework are
developed by an analysis of the proposed manufacturing strategy.
The collection of decisions determined by the production!
distribution, production process, and inventory models serves to
define, at an appropriate level of detail, the design of a manufacturing
system appropriate for that strategy.
The models also generate predictions of the manufacturing

system performance. Such performance could be measured in terms
of the fully allocated production and distribution costs by product
and market area, the level of customer service, the quality levels to be
achieved, and the flexibility of the system to changes in production
requirements. As noted in our paradigm of manufacturing strategy,
the targeted response in terms of these measures, for each product
segment, constitutes the firm's overall manufacturing mission.
The response of the firm's customers to the achieved levels of
operations performance is, of course, a fundamental determinant of
the success of the manufacturing strategy. Given consumer preferen-
ces for cost, quality, and service, repeat buy behavior will be affected
by the implementation of the manufacturing strategy. This behavior
will in turn affect the level of demand for the firm's products. These
changes take place in an environment in which other firms who bring
the same, or sufficiently similar, products to the market place with
possibly different levels of performance, are competing for the
customers' business. Market response, as measured by changes in
market share and/or the overall level of demand for the firm's
outputs, will therefore be determined by the competitive equilibrium
which results from the set of manufacturing strategies adopted by
each firm. This equilibrium will, in turn, be influenced by the
competitive structure of the industry. Various classes of markets
(perfect competition, oligopoly) and response patterns (price, new
entrants, new products integration) specific to the industry should be
considered. These responses can be fed back into the MPSS model
hierarchy as modified demand and/or price forecasts. Analysis based
on equilibrium levels of demand allows us to predict the cost and
benefits of the proposed manufacturing strategy in a manner that
takes into account the response of the firm's competitors.
5. IMPLEMENTATION CONSIDERATIONS
As noted earlier, at its current stage of development the MPSS

consists of two model systems: the production/distribution model
(PILOT) and the production process evaluation model (PRIME).
The capabilities of the two prototype software systems include
1. Evaluation of focus vs. complexity tradeoffs at the plant and

manufacturing process level;
2. Linkage of manufacturing and distribution planning systems;
3. Evaluation of supply chain economics including the impact of
vertical integration, supplier sourcing, processing technology,
and distribution design and control;
4. Evaluation of the impact of congestion, scale, and scope;
5. Accounting for the costs and benefits of flexibility m
manufacturing.
The software system developed for MPSS is illustrated in Figure
10. Commercially available linear programming, integer program-
ming, and transportation problem codes were linked in a general
solution algorithm. As noted above, PILOT must solve large-scale,
nonlinear, mixed integer mathematical programs. The optimization
models embodied in the ~(q) functions of PRIME are nonlinear
objective function, linearly constrained mathematical programs be-
cause of our scope/complexity multiplier model (here applied to
~tk)' The user interface for PRIME was written in PASCAL.
At its current stage of development PILOT and PRIME have
been successfully used on a number of large-scale problems. User
interface and report generation software is under development. The
prototype code has recently been converted to production code.
While the above list of MPSS capabilities is comprehensive, it is
clear that it does not fully address all of the issues raised in the
manufacturing strategy paradigm and in our description of the
complete set of MPSS modules. At the present time MPSS provides
for more comprehensive analysis of manufacturing strategies than
those which are attainable with models and existing software. The
MPSS software also provides near optimal solutions with very low
computer costs for very large problems. Much additional develop-
ment work and testing remains to be done. It is the intent of MPSS
and its various modules to provide analytic support and quantitative
inputs to what must always remain a complex exercise in judgement.
Models and systems can only serve as one of many informational
inputs to the manufacturing strategy analysis process.
~
-.;~~;..'":-..{:T~-;i.'~;;
..... ~
;:?: PRODUCTIONI:C c:::
:~:: DISTRIBUTION ~.:-: ~
...,(")
ll<l...
.----1.~I~\t1~E~~~;ilg·~·~);· 'f!-~:::============~~::::~~=====------- ...
~.::-~~~::;:r.(o:;!;:;_ ..::~~::":.\~,::::"::~. ~
I DISTRIBUTION
N SYSTEM
~
()
p DECOMPOSITION
U PRODUCT MASTER
T ASSIGNMENT PROGRAM
AND VENDOR
SELECTION
P G
R E
o TRANSPORTATION I INTEGER N
C E
E ROUTINE PROGRAM
R
S LINEAR A
S PROGRAM T
o 0
R R
.:J~::: ::~ ~ ~.::~ :i:~. ::;.:J:'; -;~. ~.::;:." ."I.~

.
:,,:, PRODUCTlON/ ·'\'
?i/:~ PROCESS ,~, ,}i~.
~') TECHNOLOGY ::-,:,.;
EXT. EDIToJ
[ & P. C:s '~}t!~~~~·~~!~~~~~:j,
FIGURE 10
MPSS design. ......
00
v.
ACKNOWLEDGMENT
This chapter reports on the results of an on-going research

project in Manufacturing Strategy. Partial support for this project
was provided by Booz-Allen, & Hamilton, Inc., as a part of the
Manufacturing Planning Support System Development Project. We
would like to acknowledge the assistance of the many consultants
within Booz-Allen & Hamilton who participated in the project and
who contributed their time, ideas, and comments. We would in
particular like to thank Steve Griffiths, Tom Jones, Robert Mayer,
and Mike Webber for their participation and active collaboration in
all phases of the project. The project team at Wharton included
Arunabha Chatterjee, Michael Chang, Steven Hanks, Kongkiat
Opaswongkarn, Siwan Kim, and Monique Spielberg. The efforts of
all participants in this project are gratefully acknowledged. The
authors, of course, retain all responsibilities for any errors or
omISSIOns.
REFERENCES
Buijs,]. (1979). Strategic planning and product innovation-some systematic

approaches, Long Range Planning 12, 23-34.
Chandler, A. D. (1977). The Visible Hand: The Managerial Revolution in
American Business, Harvard University Press, Boston.
Chatterjee, A., Cohen, M. A., Maxwell, W., and Miller, L. (1983).
Manufacturing Flexibility: Model and Measurements, Department of
Decision Sciences working paper, The Wharton School, University of
Pennsylvania, Philadelphia.
Cohen, M. A., and Lee, H. L. (1984). Manufacturing and Distribution
Systems Design: A Decomposition Planning Model, Decision Sciences
Department, Working Paper #84-10-03, The Wharton School,
University of Pennsylvania, Philadelphia.
Dierdonck, R. V., and Miller, J. G. (1980). Designing production planning
and control systems,j. Oper. Manage. 1(1) 37-46.
Galbraith, ]. R., and Schoemaker, P. ]. H. (1978). Technology and
organization design: A managerial assessment, in Studies in Operations
Management, (A. C. Hax, ed.), North-Holland, Amsterdam.
Geoffrion, A. M., and Graves, G. W. (1974). Multicommodity distribution
system design by benders decomposition, Manage. Sci. 20(5), 822-844.

Geoffrion, A. M. Graves, G. W., and Lee, S. J. (1978). Strategic distribution
system planning: A status report, Chap. 7 in Studies in Operations
Management, (A. C. Hax, ed.), North-Holland, Amsterdam.
Gerwin, D., and Tarondeau, J. C. (1982). Case studies of computer
integrated manufacturing systems: A view of uncertainty and innovation
processes,]. Oper. Manage. 2(2), 87-99.
Goldhar, J. P., and Jellinek, M. (1983). Plan for economics of scope,
Harv. Bus. Rev. 61(6), 141-148.
Hall, R. W., and Vollman, T. E. (1978). Planning your material require-
ments, Harv. Bus. Rev. 56(5),105-112.
Hayes, R. H., and Schmenner, R. W. (1978). How should you organize
manufacturing? Harv. Bus. Rev. 56(1), 105-118.
Hayes, R. H., and Wheelwright, S. C. (1979a). Link manufacturing process
and product life cycles, Harv. Bus. Rev. 57(1),133-140.
Hayes, R. H., and Wheelwright, S. C. (1979b). Dynamics of process-plant
life cycles, Harv. Bus. Rev., 57(2), 127-136.
Heskett, J. L. (1977). Logistics-essential to strategy, Harv. Bus. Rev.
55(6), 85-96.
Hobbs, J. M., and Heany, D. F. (1977). Coupling strategy to operating
plans, Harv. Bus. Rev. 55(3), 119, 126.
Hopkins, D. S. (1977). New emphasis in product planning and strategy
development, Ind. Market. Manage. 6,410-419.
Miller, J. G. (1981). Fit production systems to the task, Har. Bus. Rev.
59(1),145-154.
Panzer, J. c., and Willig, R. O. (1981). Economies of scope, Am. Econ. Rev.
May, 268-272.
Porter, M. E. (1980). Competitive Strategy: Techniques for Analyzing
Industries and Competitors, The Free Press, New York.
Scherer, F. M., Beckenstein, A., Kanfer, E., and Murphy, R. D. (1975). The
Economics of Multi-Plant Operation: An International Comparisons
Study, Harvard University Press, Cambridge.
Schmenner, R. W. (1976). Before you build a big factory, Harv. Bus. Rev.
54(4), 100-104.
Schmenner, R. W. (1982). Multiplant manufacturing strategies among the
Fortune 500, j. Oper. Manage. 2(2), 77-86.
Skinner, W. (1974). The focused factory, Harv. Bus. Rev. 52(3) 113-
12l.
Skinner, W. (1978). Manufacturing in the Corporate Strategy, Wiley, New
York.
Sommers, W. P. (1979). Product Development: New Approaches in the 1980s,

Booz-Allen & Hamilton, New York.
Spence, A. M. (1977). Entry, capacity, investment and oligopolistic pricing,
Bell]. Econ. Manage. Sci. 8(2), 534-544.
Stecke, K. E., Dostier, D., Browne, J., Sethi, S. P., and Rathnull, K. (1984).
Classification of Flexible Manufacturing Systems: Evoluting Towards the
Automated Factory, Working Paper, Graduate School of Business
Administration, University of Michigan.
Thompson, H., and Paris, M. (1982). The changing face of manufacturing
technology,]. Bus. Strategy Summer, 45-52.
Williamson, O. E. (1975). Markets and Hierarchies: Analysis and Antitrust
Implications, The Free Press, New York.
CHAPTER 6
THE EXECUTIVE DILEMMA

New Industrial Automation Systems
Christopher S. Fuselier
One of the most controversial aspects of factory automation

concerns the economic aspects of advanced manufacturing technolo-
gies. My experience has convinced me that this financial issue is
causing some real problems for executives. These new technologies
have a huge potential to increase productivity, but that potential is
being roadblocked in many cases by the misuse of financial analysis
tools. The problem is that we are not sure how to calculate the
returns on the investments in flexible automation.
A recent Boston university school of management survey of
management attitudes toward factory automation found that 78% of
the respondents felt that, "Most businesses in the United States will
remain so tied to traditional quantitative investment criteria that they
will be unable to realistically evaluate the potential value of
computer-aided manufacturing options." In other words, because
businessmen do not know how to run the numbers using traditional
formulas, they are procrastinating and they are not putting these
technologies to work. That same survey also reported that 92% of
the respondents believe that the biggest problem in using new
manufacturing technologies are managerial, not technical.
A recent issue of High Technology magazine commented,
"Flexible automation systems radically alter the economics of auto-
Christopher s. Fuselier. General Electric Company, Industrial Automation Systems

Department, Albany, New York 12211.
189
190 CHRISTOPHER S. FUSELIER
mation." The true dilemma is that the economics apparently have not
been explained well enough for most executives to act upon.
What is urgently needed are new cost/benefit formulas and
measurements, new manufacturing economics that go beyond the
usual return on investment evaluations to take into account the total
impact of automation on the business. These technologies do not just
change the factory floor, they change the entire business in new and
important ways.
A recent A. D. Little IMPACT report (Young and Airey, 1983)
gave an excellent example of this. The report cited a major Japanese
manufacturer which had installed a flexible manufacturing system
which enabled them to reduce the number of machines in one facility
from 68 to 18; reduce the number of employees from 215 to 12;
reduce space requirements from 103,000 square feet to 30,000; and
reduce processing time from 35 days to a day and a half. However,
the financial return over their first two years of systems operation
was only $6.9 million on an investment of $18 million.
Now, on the basis of conventional accounting principles, this
kind of return on investment would be difficult, if not impossible, to
justify despite the numerous long-term benefits to the business
mentioned above.
The point is, many manufacturers continue to make the mistake
of looking at FMS investments as tactical decisions with only
short-term financial implications, when in fact they are strategic
decisions, with far-reaching competitive implications. Automation
provides permanent long-term savings and a host of intangible
benefits, such as higher quality, shorter production runs, increased
worker satisfaction, faster responses to changes in market demand,
quicker throughput, greater flexibility, and so on.
One of the most common errors managers make in evaluating
these decisions is focusing on direct labor as the target of automation.
In the past, direct labor accounted for as much as 50% of total
product cost, but today, thanks to great effort to reduce it, direct
labor accounts for an average of only 10% of manufacturing costs.
Material costs are 55%, indirect labor is 15%, and overhead costs
total 20%. Therefore, it is better to concentrate on material,
inventory, and overhead costs.
Another mistake is considering individual pieces of automation
THE EXECUTIVE DILEMMA 191
equipment in isolation from the rest of the process. In fact other

resulting benefits must be considered.
Using a robot to illustrate the problems with traditional ROI
analysis, the old way of justifying this equipment is to consider the
direct labor savings. This is a mistake because direct labor is only
10% of product cost. The other problem is that the robot is often
looked at in isolation from the rest of the plant. What good is
increased throughput here if the rest of the factory cannot handle it?
The robot may only move the bottleneck someplace downstream.
What is more, many people fail to take into consideration the
inventory leverage of a piece of automation equipment like a robot. If
you can increase the amount of time the inventory is being worked
on by the robot, and other work stations down the line, then you
have tremendous potential to reduce material and overhead costs.
Another pitfall is the failure to consider the positive effects of
improved product quality on the productivity of the entire business.
Many automation projects in GE are being justified not only on
direct productivity savings but on reduced warranty and scrap and
rework costs that result from improved quality. Another major
justification is the increase in market share and sales that result from
higher-quality products. Fortunately, we have found in our automa-
tion activities that quality and productivity are inseparable.
The fact is, without quality, you cannot have high productivity.
A commonly accepted formula says productivity equals output
divided by input. All of us would agree that if additional material
must be input to replace that which is scrapped in the process of
manufacture, or if additional labor must be input to rework products
that have not been manufactured properly, the denominator of this
equation increases, while the numerator stays constant. The result is
that productivity goes down.
The most fundamental definition of quality is conformance to
requirements, where the requirements, are established by the
marketplace. After all, the customer defines and is the final judge of
quality. The easiest way to improve the productivity of a business is
to make a quality product the first time. That reduces the denomina-
tor in our equation and causes productivity to go up. What is more,
you create a powerful new equation in the process: productivity plus
quality equals competitiveness. You also discover that improved
quality provides tremendous leverage for the business by increasing

your ability to compete.
Where does automation fit into this equation? A well-planned
and well-executed automation system is a key that can unlock both
doors to productivity and quality. It goes without saying that
automation can increase output, but automated equipment is also the
most disciplined, and therefore the toughest quality inspector you
can have. Quite simply, automation affords you the opportunity to
design and manufacture products that conform to requirements every
time. That is because it has certain attributes which can be exploited
in the factory such as consistency, repeatability, accuracy, and
responSIveness.
Automation is not a quick fix for quality problems. Installing
robots or computers without first establishing the proper disciplines
can be disastrous, because those robots will only enable you to make
more bad products faster. This brings us to the heart of the
justification issue and the problem with traditional analyses which
focus on the payback of the specific factory floor application. In fact,
automation touches more than manufacturing. It affects three impor-
tant dimensions of the business, namely, engineering, where you
build in quality; manufacturing, where lower costs result from having
designed a simpler product; and information management, where you
monitor and control the quality and flow of the entire business in
real time.
Justification of these advanced technologies has to be made on
the basis of benefits in all of these disciplines, not just direct savings
in manufacturing alone. Also the automation plan must consider
these three areas to keep the FMS from just moving a bottleneck
from one point to another.
The first project I would like to present as an illustrative case
concerns the G.E. Laundry and Dishwasher Products Division. The
project scope included modernizing and automating the 30-year-old
dishwasher plant in Louisville, Kentucky. A total of $38 million was
spent over the four-year period from 1979 to 1983 on this project.
The goal of the project was to bring the plant from a
semiautomated, traditional assembly operation to a completely mod-
ernized CAD-CAE supported factory management system. This
included closed loop machining with fully automated materials

handling capabilities to allow flexible single-lot manufacturing of
individual products. The materials handling system consisted of
flexible transfer lines, coordinated with automated parts picking
systems. Extensive use of robotics substantially automated the
assembly and testing operation. The result was an integrated flexible
manufacturing system capable of quick changeovers, small lot sizes,
and high-quality standard production.
The economic and productivity results achieved in the Louisville
plant were gratifying. By the end of 1983 the following statistics had
achieved steady state relative to the pre-FMS application:
• 10% reduction in total cost of labor, material, and overhead;

• 33% reduction in cycle time;
• 40% reduction in scrap and rework;
• 25% increase in productivity;
• 78% fewer parts, which significantly reduced inventory and
handling costs.
Based on the success of this project, G.E. recently committed to

spend over $800 million automating other major appliance
businesses.
The next case I would like to report on concerns the G.E.
Transportation Systems Division. The project included the use of
sophisticated computerized information systems for routing and
control of parts, tools, and assembly. The flexible machine system
was centrally controlled and operated through a local network of
numerically controlled machines and robots. Local programmable
control in certain high-technology centers was also included. This
was all coupled with a fully automated heavy-duty transfer line for
materials handling. The results were as follows:
• 33% increase in product capability;

• 20% material savings;
• Labor savings;
• Improved product quality;
• 33% reduction in assembly cycle time;
• 20% increase in inventory turns.
The technologies of computer-aided design and computer-aided

engineering now not only allow us to visualize the full, detailed,
three-dimensional design of an object to be manufactured, but also
provide the capability to measure the performance of the object in its
destined environment, and to assess the manufacturability of the
object of automated techniques. With these advanced tools you can
build, test, and analyze a computer prototype with a sense of realism
that approaches working with the actual product.
These new engineering design tools play a critical role in the
justification equation, because they can reduce the time it takes to
bring the product to market, reduce material costs, and assure
manufacturability before parts are produced in volume.
Manufacturing automation is the set of capabilities now available
for the factory floor that improves the way materials are processed
and finished. Numerical controls, programmable controls, robotics,
vision systems, and automated material handling or retrieval systems
are all examples of current technology in manufacturing automation.
By utilizing these technologies in combination, levels of process
control, and process yields never before experienced can be achieved.
The third element of factory automation-information
management-provides the automated tools to monitor and control
the quality of the entire manufacturing process on a real time basis. If
problems occur, decisions are made and communicated through local
area networks, programmable controls, numerical controls, robot
controls and other intelligent devices to reconfigure the flow of
materials through the factory, automatically make process ad-
justments, isolate any problem areas, and optimize their repair and
maintenance.
At GE, we try to get our businesses to consider the parallel
implementation of an information strategy and a material flow
strategy. By that, we mean looking at everything that goes on in the
business process as an impediment in the flow of material and data
from the receiving dock to the shipping dock. Until a business is able
to look at its factory as a flow shop, rather than a job shop, it will
continue to have a fragmented manufacturing strategy and a piece-
meal justification.
In summary, to fully understand the financial implications of

investing in flexible, programmable manufacturing technologies, you
must consider the three major elements: engineering, manufacturing,
and information as an inseparable, integrated whole. You have to
consider all of the elements of product cost: material, overhead, and
indirect labor-and keep reminding yourself that direct labor is the
smallest piece of the pie. You must also recognize the pervasive
impact of improved product quality on both product cost and market
share. At General Electric we believe strongly that a systems
approach to automation is the only way to get at the financial
justification issue that we have been talking about. Technology alone
is not the answer. Businessmen must avoid looking at each piece of
equipment in isolation from the rest of the company. The key is
fitting these technologies together with appropriate management
disciplines, inventory disciplines, and process disciplines. That is the
key to making the total payback greater than the sum of its parts.
REFERENCE
Young, c., and Airey, J. (1985). "Economic Justification for Flexible

Manufacturing Systems: Counting the Strategic Benefits," A. D. Little
IMPACT report.
PART III
PRODUCTIVITY MANAGEMENT
CHAPTER 7
WORKING SMARTER
THE JAPANESE WAY
The X -Efficiency of Theory Z Management
John F. Tomer
1. INTRODUCTION
"The amount of money that they [American automobile

industry] are spending doesn't bother me-capital investment
alone will not make a difference .... When Detroit changes its
management system, we'll see more powerful American
competitors. "
Executive Vice President of Honda as quoted in Grayson
(1982).
A number of recent organizational behavior studies (see, for example,

Ouchi, 1981; Pascale and Athos, 1981) agree that the important
differences between Japanese management and American man-
agement explain in large part why Japan, and not just its auto
industry, has been able to achieve rates of growth in productivity
much higher than the U.S. since the 1950s. These studies suggest that
a critical dimension of Japanese management is the nature of their
employer-employee relationships which motivate organization
members to work "smarter" than their counterparts in the U.S. In
other words, Japanese firms are simply more X-efficient, i.e., their
productivity is closer to potential, than American ones. If so, why?
John F. Tomer. Department of Economics and Finance, Manhattan College,

Riverdale, New York 10471.
199
200 JOHN F. TOMER
The purpose of this article is to adapt economic theory (notably

X-efficiency theory and transaction cost economics!) to the task of
explaining why one would expect organizations managed Japanese
style to be more effective in achieving both labor productivity closer
to potential (for given technology and capital inputs) and growth in
productivity.2
It is very clear that over the last several decades the growth of
Japanese productivity has been much higher than that of the u.s. and
significantly higher even than the fastest growing developed econo-
mies. For example, during 1960-1981, japan's growth in labor
productivity in the manufacturing sector averaged 9.2% compared to
2.6% for the U.S. 3 For all industries during 1960-1976, the growth in
japan's labor productivity was 7.5% compared to 1.7% for the U.S. 4
A number of empirical studies have focused on the sources of
Japanese economic growth in comparison to those of the u.s.
However, these "aggregate explanations" of economic growth leave
much to be desired and usually are only suggestive about the role of
management. Edward Denison and William Chung (1976, p. 83), for
example, indicate their belief that "management is a key element in
efficiency," but they are unable to quantify this factor; as a result,
changes in managerial performance are reflected largely in their
residual, the portion of economic growth unexplained by the
measured input changes. Noteworthy is Denison and Chung's (1976,
p. 82) observation that Japan has been able not only to adopt new
knowledge rapidly but seemingly to catch up to "best practice,"
whereas the gap between best and typical practice is not closing in
other countries. 5 Could the nature of Japanese management be the
key ingredient that enables Japanese firms to achieve high knowledge
utilization ?6
To facilitate the analysis, the discussion focuses on the essence of
the Japanese management ideal taken out of the cultural context of
Japan. This ideal is called "Theory Z" or simply Z management,
following the usage of William Ouchi, while American or Western
style management is referred to as A management. 7 The essential
characteristics of Z management are as follows: 8
1. Lifetime employment ideal is effective in practice.

WORKING SMARTER THE JAPANESE WAY 201
2. Strong corporate culture and philosophy (or superordinate

goals) provide for implicit control of employees who in-
ternalize these values in contrast to A management's emphasis
on explicit control mechanisms. 9
3. Community or family relations characterize relations among
employees, i.e., wholistic rather than pure economic relations
predominate.
4. Decision making and responsibility taking are collective in
nature; the decision-making process emphasizes consensus
formation.
5. Careers are characterized by slow evaluation and promotion
and nonspecializatoin.
The operations of Japanese firms reflect the Z management ideal

to quite different extents. It is in the largest Japanese firms where the
above characteristics are most clear-cut and distinctive. Therefore, to
put the analysis in proper perspective, it is important to discuss a
number of the key features of the Japanese industrial setting.
Japanese companies, which tend to be narrowly specialized,
engaging in one line of business or a few closely related ones (Clark,
1979, p. 49), are customarily thought of as being arranged in a
hierarchy according to well-established criteria. Companies achieve
their status in this "society of industry" according to "the business
they are in, their size and market share, their affiliations with bigger
companies and banks, the level of wages they pay, their methods of
recruitment, their customers and suppliers, and their shareholders"
(Clark, p. 95). Shareholding in Japan, unlike many Western coun-
tries, is largely an expression of the relationship between institutions
holding shares in each other rather than an expression of individual
financial interest. Thus, firms are able to subordinate the profit goal
to the goals of service to society and gaining higher status in the
society of industry (pp. 136-137, 95-97). Small companies supplying
the largest firms are often the ones on the lower rungs of the
hierarchy; these suppliers are generally less financially secure and
more dependent on and dominated by their larger brethren. While
this company "dualism" in Japan is widely cited, it is not uncommon
for small companies to grow and, in the transformation process, take
202 JOHN F. TOMER
on the characteristics (including higher productivity) of the higher

status group (pp. 94-95).
The dualism above is very much reflected in differences (among
firms and within firms) in employer-employee relations and the
workings of the labor market. A "lifetime" of employment in a work
community is clearly a widespread ideal, and companies which are
substantial enough to make their promises of lifetime employment
credible do practice it (Clark, 1979, pp. 79, 174-179). The best
companies recruit labor from the best universities by offering an
attractive "package" of financial and nonfinancial compensation over
a lifetime which keeps turnover low. Thus, in Japan the secondary
labor market, relating to movements of labor between firms, is
secondary both in the sense of its quantitative importance and in the
sense that this labor is utilized more by lower status firms (Chapter
5). The best companies generally have more formal pay and
promotion standards, and nonmerit factors such as age, seniority, and
family size are more important criteria in these decisions than for
lower-ranking firms (pp. 45-46, 112-125). It should be noted that
even many of the best companies have substantial groups of workers
(e.g., temporary workers and women) who do not share fully in the
financial and intangible benefits of company membership (p. 234).
Also notable is that unions in Japanese companies are enterprise
unions, not unions organized for an entire industry or nation, and,
though they look after employee interests, they are remarkably loyal
to and identified with their companies (pp. 45-46).
The important question is: Are the Japanese firms which are
managed Z style economically successful in spite of or because of
these characteristics? Some observers have maintained that these
attributes are leftovers from a feudal past and bound to change as
modernization progresses (see, for example, Marsh and Mannari,
1976). Or, are they the key to superior performance? My thesis is
that the essential features of Z management make possible greater
X-efficiency than does A management.
1.1. X -£fficiency and Productivity

To put the following analysis in perspective, it is important to be
clear about the relationship between X-efficiency (XE) and produc-
tivity. XE theory takes X-inefficiency, productivity less than poten-

tial, to be the usual state of affairs. X-inefficiency implies that costs
are above minimum costs and that changes in the XE factors could
lower costs. The XE factors may be (1) external to the firm, for
example, the state of competition, or (2) internal to the organization.
The latter reflects the firm's investment in organizational capital
(Tomer, 1981a), namely, the extent to which it has made explicit
efforts to achieve lasting improvements in its organization
effectiveness.
The labor productivity of a firm at one point in time can be
considered to be a function of (1) its tangible capital, (2) its "hard"
(nonmanagerial) technology, (3) the attributes of its laborers (human
capital), and (4) its XE factors. The XE factors are also a chief
determinant of productivity growth over time. X-inefficiency, in a
dynamic context, means less innovation than possible, which means
less growth/improvement in both the XE and non-X-efficiency
(capital, technology, labor) factors than possible. The analysis below
seeks to show how particular XE factors (internal to the organiza-
tion) associated with Z management make possible either labor
productivity closer to potential or faster growing labor productivity
than A management does.
1.2. Plan of the Chapter

The chapter has two main sections. Section 2, which follows,
begins by extending the notion of implicit contract (a nonlegally
binding understanding between employer and employee) to one of
implicit psychological contract. Firms may be located along a
spectrum of implicit psychological contracts according to the nature
of their management; Z-managed firms are found on the far right of
the spectrum. Perspectives from Oliver Williamson's transaction cost
economics, especially the concepts of opportunism and bounded
rationality, are found to be useful in explaining why organizations
located on the right of the spectrum should have lower costs.
The point of departure for Section 3 is X-efficiency analysis a la
Harvey Leibenstein. His effort determination analysis is modified
and extended to enable Z and A organizations to be compared with
respect to worker motivation. The central question is: Why is it
204 JOHN F. TOMER
rational for Z workers to exert higher effort than their counterparts

in A organizations?
2. IMPLICIT PSYCHOLOGICAL CONTRACTS,

TRANSACTION COSTS, AND COOPERATION
2.1. The Implicit Psychological Contract

The essence of the relationship between the employee and his
organization is described by the evolving implicit psychological
contract (IPq between them. The IPC concept, on the one hand, is
an extension of the notion of employer-employee implicit contracts
developed by Arthur Okun (for example, 1981) and others. Such
implicit contracts involve nonlegally binding understandings relating
to compensation, work conditions, and the nature of work to be
done. While more flexible than explicit contracts, implicit contracts
have the power to command at least a minimum acceptable level of
fulfillment; otherwise one of the parties to the contract will be
considered to be in default and the continued existence of the
relationship will be called into question by the other (dissatisfied)
party. In the usage of both Okun and Leibenstein (for example, 1976,
p. 146), implicit contracts relate largely to relatively tangible,
economic concerns such as wages, hours, work breaks, job safety,
and other conditions. The IPC, on the other hand, is much the same
as Chris Argyris' (1960) "psychological work contract," Harry
Levinson's "psychological or unwritten contract," or John
Gabbarro's (1979) "interpersonal contract" (see also Thomas, 1978).
That is, "it relates to a set of mutual expectations concerning
performance, roles, trust and influence ... ;" it specifies " ... what
each should contribute to a relationship and what each should get out
of it" (Gabarro, 1979, p. 10).
It is useful to think of a spectrum of implicit psychological
contracts (Figure 1). On the far left of the spectrum, the relationship
between employer and employee is limited and of short-term
duration; therefore the implicit contract aspect of the relationship is
negligible since the critical elements of the relationship are considered
Short Term Intermediate or Long or "lifetime"

- - - - - Indefinite Term - - - - - Duration
Discrete Transaction only _ _ Some Rules and Conventions _ _ Community

for Fair Play
Employee Contribution Commitment of Employee

Limited to Economic Service - - - - - - - - Far Beyond
Economic Contribution
Employee as Agent of Principal Employee Shares Goals/

- - - - - - - Shares in Decision Making
Control through Explicit Control Implicit Control through

Explicit Contract--- (Directing, Monitoring, - - Internalization of Goals
Rewards, Penalties) and Values
_ _ _ _ Autocratic _ _ _ _ _ _ Supportive Leadership,

Participative
Theory X _ _ _ _ _ _ Theory Y _ _ _ _ _ _ Theory Z
FIGURE 1
Implicit psychological contract spectrum.
by the explicit contract specifying the worker's contribution and

compensation. In the middle of the spectrum are the intermediate or
indefinite-term relationships, where, according to Okun (1981, pp.
84-85), the "invisible handshake" involves rules and conventions for
fair play that curb distrust. On the spectrum's far right, the
relationship is for a work lifetime. In return for an implicit guarantee
of lifetime security and community, the organization demands from
the employee an almost total commitment-not only a technical or
professional contribution but an emotional, even spiritual, one.
Another important variable defining the IPe spectrum is the
nature of the principal-agent relationship. On the left side of the
spectrum, it is very clear that the employee is an agent of the
organization's principals; the employee is expected to try to achieve
206 JOHN F. TOMER
what the principals decide is consistent with their goals. Moving

rightward past the middle of the spectrum, the principal-agent aspect
of the relationship begins to fade until, on the far right, the employee
can no longer be considered to be an agent of the organization; at
this point, he/she is of the community, shares its goals, and
participates in its decision making. 10
The predominate method of organizational control is another
variable defining the IPC spectrum. On the far left, the explicit
contract provides all or most of the control. Explicit control, which
involves directing, monitoring, rewarding, and penalizing behavior,
begins on the left side and fades a little to the right of center. On the
right, where employees share the organization's superordinate goals
and values, control by the implicit mechanism of shared goals
predominates. The movement from the left end of the explicit control
range to the far right can be characterized as moving from Theory X
(autocratic) through Theory Y (supportive leadership, participation)
to Theory Z (see Maslow, 1971, pp. 274-277). Typically, A
management is near the center,u
Why is lifetime employment, perhaps, the key ingredient in Z
management? My analysis suggests that when a firm decides to offer
lifetime employment, it has a great incentive to develop an IPe
which includes most of the other characteristics associated with the
spectrum's extreme right. After all, how can an employee or an
employer make a narrow economic-legal contract with a lifetime
duration? The firm, for example, would be unable to predict a
lifetime's worth of demand for particular products, much less
demand for particular labor skills. A worker, on the other hand, will
not generally be able to predict the evolution of his work preferences
and abilities far in advance. Thus, when lifetime employment is
involved, the understandings must be in the form of broad, enduring
principles which guide and inspire the relationship. Not surprisingly,
these lifetime relationships are generally characterized by trust and
flexibility.
2.2. The fPC Hypothesis
My hypothesis is that companies whose IPCs are more right-

ward on the spectrum will be able to achieve lower costs than other
companies, other things being equal. 12 Two key reasons are, as

Oliver Williamson's transaction cost economics suggests, (1) oppor-
tunism and (2) bounded rationality. Opportunism is not simply
self-interest seeking, but self-interest seeking with guile. An example
is where a party engaged in bargaining makes false and misleading
(self-disbelieved) statements in order to gain at the expense of others
(Williamson, 1979, pp. 234, 241-2). Bounded rationality refers to the
distinct limits of human mental ability when it comes to performing
complex calculations or planning for an uncertain future.
To appreciate how opportunism and bounded rationality are
related to the IPC hypothesis, it is necessary to consider the essence
of the transaction cost economics perspective. In the view of
Williamson and Ronald Coase (1937), it only makes sense to organize
an extra transaction or activity within the firm when the costs
associated with carrying out the transaction through market exchange
or organizing in another firm are equal or higher. The market
alternative will be higher cost than internal organization (or hier-
archy) when the costs of making mutually satisfactory explicit
contracts (i.e., transaction costs) are relatively high. The latter are
likely to stem from (1) the need to consider many complex
contingencies and (2) the necessity to renegotiate contracts due to
unanticipated events. The more one suspects the other party of
opportunism, the more one is likely to devote additional resources to
contract making. This is especially true when uncertain future events
create the opportunity for one party to gain at the expense of the
other. In Williamson's view, hierarchy will be utilized to the extent
that the internal relationships (or implicit contracts) among parties in
the firm act to attenuate opportunism (substituting a greater element
of trust) and make relationships flexible enough so that neither
complete anticipation of complex contingencies nor periodic renego-
tiation are necessary. Williamson has utilized this perspective in
studying product markets; in what follows I have utilized certain of
its features in analyzing labor markets, or, more specifically, the
relationship between employer and employee.
One reason why firms with rightward IPCs are hypothesized to
have lower costs is that they are likely to experience lower labor
market transaction costs, i.e., the costs associated with contracting for
labor. Why? Rightward IPCs (characterized by shared values,
208 JOHN F. TOMER
community, and trust) act to attenuate opportunism. As a result,

there is less need (1) for contracts protecting against the distrusted
party, (2) for contractual provisions dealing with complex anticipated
contingencies,13 and (3) for renegotiation arising from unanticipated
events. Moreover, the long-term mutual commitments involved in
rightward IPes mean fewer resources are required for firing, hiring,
and training of employees with the termination or start of individual
labor contracts. Since less resources are required for these activities,
costs, namely, labor market transaction costs, will be lower. It should
be noted, however, that the organizational change required to
develop more rightward IPes may require significant resources.
These organizational changes as well as other types have frequently
been the occasion for large expenditures on management consulting
services. Those expenditures along with the output foregone when
the firm's employees direct their efforts to organizational change (i.e.,
the opportunity costs) comprise the total organizational investment
cost (see Tomer, 1981a, for an empirical analysis). Like any in-
vestment it will be justified if the lowered future costs are large
enough relative to the magnitude of this investment in organizational
capital considering the risks involved.
The implicit psychological contract concept may also be applied
to the relationships among employees in a firm as Levinson (1963, p.
37) has done. Here the IPe refers to the mutual expectations among
individual employees and among groups of employees, i.e., to the
psychological quality of these relationships. We would expect that
these IPes would be influenced or conditioned by the IPe between
employer and employees; nevertheless, employees do exercise sig-
nificant discretion with respect to the relationships they develop with
others in the organization. Again, it is useful to think of a spectrum
of these IPes. On the right side, the IPes are characterized by much
interdependence, communication, and cooperation. On the left, the
employee's formal relation to the firm (his specialization, the
technology used, and explicit contract) dominates his work relation-
ship with others. Thus, there is a relative lack of interaction,
communication, and cooperation with others on work-related mat-
ters and relatively more independence. Clearly Z firms are to the
right of A firms on this spectrum, although how much further may
be a matter for debate.
This brings us to the second reason why firms with rightward

IPCs are likely to experience lower costs. Rightward IPCs
(employer-employee) provide a more favorable environment for
rightward IPCs among employees, and the latter involves more
cooperation. As it turns out, the degree of cooperation is an
important determinant of productivity. Let us consider why.
Cooperation refers to mutually helpful behavior among workers
in an organization. The need for this reciprocating behavior arises
when organizations have either tasks too large for individual ac-
complishment or tasks with elements requiring quite different
specialized abilities. Ordinarily, one of the first steps in doing a task
is to divide it into sub tasks and to assign these to individual workers
who are evaluated and rewarded according to their achievement on
these subtasks. It is, however, difficult in many cases to anticipate all
the necessary subtasks and the quantity and quality of labor required
for each of the subtasks. This difficulty is, of course, a manifestation
of bounded rationality. If cooperative behavior prevails, the unanti-
cipated work gets done as some employees are willing to do helping
work for which the credit is likely to accrue to others (the employees
being helped). One of the reasons why these employees are presum-
ably willing to help is because they anticipate their own need for
future help, and thus expect similar reciprocation from their fellow
workers. This cooperative pattern implies mutual positive in-
teremployee externalities. That is, employee B's helpful actions lead
to benefits for employee C (higher evaluation, more positive recogni-
tion, feelings of job success, etc.). Employee B may not realize any of
these benefits from his actions, but is likely to later when C helps
him.14
Noncooperative behavioral patterns among workers may, of
course, exist. In a "zero" cooperation pattern, no helping occurs, i.e.,
no positive externalities. Another possibility is a sabotaging (negative
cooperation) pattern. Here, B's actions are intended to undermine
C's chances for success on C's subtask, and vice versa. Presumably,
this negative externality pattern occurs when both Band C believe
their opportunism will be rewarded despite its effects on the
organization. IS
The costs of these noncooperative patterns are clear. First, there
is the output foregone. The zero cooperation pattern involves tasks
210 JOHN F. TOMER
not completed due to lack of cooperation; sabotage involves

destroyed output. Certainly, management may take actions which
counter the effects of "spontaneous" noncooperation in order to
maintain output. However, these managerial actions can themselves
be costly in the sense of using up resources and are the second of the
noncooperation costs. To compensate for lack of cooperation,
managers may (1) do more complete task planning, resulting in less
unanticipated work, and (2) create new divisions of labor to
accomplish work not previously anticipated. 16 ,17
The basic point is that rightward IPCs (employer-employee)
provide the organizational environment conducive to rightward IPCs
among employees; the latter, especially, attenuate opportunism and
promote trust among employees, thereby encouraging cooperation.
On the other hand, leftward IPCs are more likely to be associated
with low or zero cooperation or even sabotage along with man-
agement efforts to reduce these effects, i.e., higher costs of
noncooperation.
3. THE DETERMINANTS OF EFFORT
3.1. The Utility-Effort Relation

The superiority of Z management is not limited to its reduction
of labor market transaction costs and noncooperation costs relative to
A management. This section argues that Z management is associated
with higher worker motivation. The starting point is Harvey
Leibenstein's X-efficiency analysis, which focuses on the effort
(directed work effort) choices of workers. Increased worker effort
may mean working more intensely, longer, with better direction,
with higher-quality output, with better coordination, or even work-
ing more innovatively, responsibly, or cooperatively. The effort level
chosen, according to Leibenstein (1976), reflects in essence a com-
promise between one's own desires and the demands of the organiza-
tion. Increased pressure involving a variety of organizational rewards
and punishments generally leads the worker to select a new higher
effort level,18 Is there any reason to believe that workers in Z
organizations will be more X-efficient and choose higher effort than

their counterparts in A organizations? No! Applying XE theory
straightforwardly leads, if anything, to the conclusion that A
organizations will be more X-efficient than Z organizations since the
former will generally make more explicit demands on workers and
apply more pressure. One should bear in mind, however, that XE
theory was originally developed for application to A organizations,
i.e., organizations where a clear principal-agent distinction between
employer and employees is involved. For the purpose of comparing
Z and A companies, XE theory requires some modification.
First, consider the essential features of Leibenstein's utility-
effort relation. An employee's current utility from work effort (V) is
a function of the effort level chosen (E), the demands of the
organization (DO) and the individual's personality (P). The relation-
ship between V and E for a worker can be depicted graphically as
curve V l in Figure 2. This worker will choose effort level, Ei, in
order to achieve maximum utility Vi. If the organization makes
increased demands upon the worker, the VE relation or curve shifts
to V 2 (rightward and possibly upward). Now the optimal effort point
is Ej as the worker accommodates to increased demands and raises
his satisfaction level above what it would have been at Ei.
The analysis of Alchian and Demsetz (1972) is parallel in a
number of ways to that of Leibenstein. They argue that the
"contractural structure" of the classical capitalist firm makes it more
U U*
2
U*
1
E* E* E
1 2
FIGURE 2
The utility versus effort relationship.
212 JOHN F. TOMER
efficient in production than decentralized contractual relations

(markets) when team production yields an output higher than
laborers working separately could produce. The key problem in team
production is that it is very costly or impossible to measure each
laborer's marginal product and to reward each accordingly. Thus, it
is necessary to monitor the behavior of each laborer to find clues
regarding their productivity. Monitoring, however, is not a costless
activity, and workers will generally find that an organization's
incentives lead them to shirk (substituting on-the-job leisure for
work effort) more than is optimal from a societal point of view. The
contractual structure of the classical capitalist firm is much the same
as what has been referred to here as A management. Thus, when the
capitalist-managers (a la Alchian and Demsetz) step up their monitor-
ing to lower shirking in response to a decline in their residual claim
on the firm's income, this is essentially the same as when
Leibenstein's principals put more pressure (increased organizational
demands) on worker-agents and, thus, expect greater effort. Alchian
and Demsetz have not applied their analysis to Z style organizational
relations, and it does not follow straightforwardly how Z man-
agement would lower shirking, i.e., raise effort, motivate workers,
etc.
My (1981b) paper proposes a useful modification of
Leibenstein's UE relationship in which a work environment (WE)
variable (including dimensions such as the nature of the implicit
contract between employees and the company) is an important
addition. With this revision, U is a function of E, P, DO, and WE.
Improvements in the work environment (for example, allowing more
worker autonomy, greater participation in decision making, im-
proved training and personal growth opportunities, and more recog-
nition of worker efforts) are expected to shift the UE curve
rightward, resulting in higher effort choice; prolonged pressure, on
the other hand, may have a destructive effect on the work environ-
ment, and thus lead to lower effort (Likert, 1961, p. 71). Insofar as Z
organizations have work environments reflecting more rightward
IPes than those of A organizations, and insofar as Z managers are
facilitators, and therefore, less likely to use "pressure" tactics, this
revised XE theory suggests why higher effort will be chosen, i.e.,
why higher worker motivation is likely in Z companies. However,

further progress seems possible.
Suppose workers' effort choices depend on the flow of utility
(current and future) expected from current work effort, i.e., Vt(Eo)
for t = 0, 1, ... , m, where Eo is current period effort. 19 Future utility
expectations of workers are based on past and present organizational
experience as well as subjective considerations. Let us assume that
workers weight (or discount) the stream of Vt(E o), i.e., they base
their E choice on the present value of Vt(E o), which we denote as V'.
Thus,
(1)
where k is the appropriate rate of discount and m is the worker's

utility time horizon taken to be closely correlated with the worker's
expected tenure in the organization. 20 I do not maintain that workers
actually do the quantitative calculation suggested by (1), but it is my
belief that they take into account expected future satisfactions
(promotion, pay increase, pride in accomplishment, pride in good
relations among fellow workers, etc.) and that V further in the future
is weighted less. In other words, workers seem to act as though they
make the calculation. 21
Assuming workers desire to maximize V', workers' E decisions
will be similar to corporate inve~tment decisions. Suppose workers'
decisions are focused on discrete increments of E, say !1E equal to
one. Workers would be expected to accept the !1E if !1V', the
resulting change in V', is positive, much like they might accept an
investment if the net present value is positive:
!1V' = for!1E = 1 (2)
While this may seem farfetched, it nevertheless provides a useful

framework for understanding the superiority of Z management.
The essence of the argument is that Z workers, as compared to A
workers, (1) have good reason to expect higher increments in current
214 JOHN F. TOMER
and future utility from extra current effort, (2) are likely to discount
future utility at a lower rate, and (3) more of their efforts are likely to
be directed toward organization goals. In other words, workers in Z
organizations are hypothesized to find the extra satisfactions afforded
by working more responsibly, cooperatively, innovatively, with more
attention to quality and organizational goals, etc., substantial and
certain enough to motivate them to adopt higher effort positions.
The main reasons for expecting higher effort outcomes in Z
organizations are outlined below. First, !1U' is more likely to be
positive for Z organizations when considering a given !1E because the
Z time horizon, mz, is generally longer than A's, mA> owing to
lifetime employment. That is, for given !1E,
mZ !1U mA !1U
!1U' - '" > '" !iU' (3)
z - ~ (1 + ky
t
;90 (1 + kY
t
-- A
since mz > m A, assuming for the moment no differences in !1 Ut or k
between Z and A organizations.
Higher increments in current and future utility from extra effort,
!iUt , in Z organizations might be expected for the following reasons.
First, the wholistic orientation of the Z work environment allows
workers, by virtue of their efforts, to satisfy their higher needs,
especially self-actualization, and enables workers to experience high
self-esteem. Second, the superordinate goals of Z companies may
enable worker efforts to be more satisfying as a result of the
inspiration and meaning they provide. Third, if as expected relation-
ships among workers in Z organizations are based more on trust,
workers can expect more future satisfaction from current cooperative
efforts as other workers reciprocate their efforts, even if with
substantial lapse of time.
Z workers might use a lower discount rate because they associate
lower personal risk with their efforts, i.e., they perceive the distribu-
tion of expected future utility outcomes to be narrower. There are
two reasons for this. One is the existence of a definite and lasting
corporate philosophy; this provides insurance that future work
accomplishments deriving from current and future efforts will not
suffer an unexpected decline in value from the organization's
standpoint. Second, "automatic" pay increases and advancement,

especially in the early work years, ensure that in the event of failure
resulting from one's efforts, the personal consequences will not be
catastrophic. Third, performance evaluations emphasize the long
term and make effective use of subtle and nonquantifiable informa-
tion. Thus, to the extent that rewards follow from performance, there
will be a more certain connection between them over the long term,
largely eliminating one of the factors that may contribute to the
spread of future utility outcomes.
There are two other possible reasons for a lower discount rate in
Z organizations. The first is that Z workers may have a lower rate of
utility time preference owing to their superior capacity to delay
gratification. The evidence from the psychological studies sum-
marized by Maital and Maital (1977, pp. 188-193) is quite suggestive
here. From these studies, it is clear that lack of trust and social
maladjustment are positively related to an individual's rate of time
preference. To the extent that organizational relationships in Z
companies are relatively more conducive to the development of trust
and positive social adjustment, this should lead to lower rates of time
preference among Z workers.22
The other argument for a lower k is that Z workers may be less
risk averse due to the greater prevalance of group decision making in
Z organizations. This phenomenon is known as the risky shift; it
"states that groups influence individual decision making toward
positions of higher risk a significantly greater number of times than
not, and under almost any conditions" (Byrd, 1974, p. 46). When
individuals working in groups are willing to accept decisions
involving higher risk for their organizations, and presumably for
themselves, without greater expected rewards, this implies lower risk
aversion, and thus, a lower discount rate.
Why do workers in Z organizations have an incentive to direct
more of their efforts toward organizational goals? For one, workers
with lifetime employment do not need to devote their energies to
"resume and professional credential maintaining" activities. Second-
ly, both automatic and collective aspects of compensation serve to
inhibit the most harmful aspects of competition between workers,
and thus, reduce the energy devoted to the type of competitive
216 JOHN F. TOMER
activity not in the organization's best interest. Third, the facilitative

management style in combination with the superordinate goals may
encourage workers to envision themselves making larger contribu-
tions to the organization than would be the case if specific work
goals were determined by managers.
Are there reasons to believe that Z organizations will reduce the
costs of shirking by reducing both shirking and the costs of
monitoring it? The analyses above which explain why Z workers will
exert more effort than A workers are at once arguments [though not
cast in the Alchian and Demsetz (1972) framework] explaining why Z
workers may shirk less. Robert Crawford (1983) considers the
shirking cost question explicitly, and others have developed related
insights. First, Alchian and Demsetz (1972, pp. 790-791) have
pointed out, almost as an aside, that "if one could enhance a common
interest in nonshirking in the guise of team loyalty or team spirit, the
team would be more efficient." The development of team spirit and
loyalty through "cultural socialization" is, thus, one way that
Japanese companies have attempted to keep shirking to a minimum
(Crawford, 1983, p. 10). Second, the common practice in Japanese
companies of paying large bonuses "proportional to the productivity
of the whole team" provides the worker an incentive to behave as a
cooperative team member, discouraging shirking within the team (p.
7). Cable and FitzRoy (1980, p. 104) add that worker's participation
in managerial functions makes them more aware ot the reliable
connection between their efforts (nonshirking) and their profit (or
bonus) share.
Monitoring costs are also likely to be lower in Z organizations.
First, these costs will be low to the extent that employers refrain
from opportunistic actions and foster mutual trust as this will
discourage employee deception related to job performance outcomes
(Crawford, 1983, p. 9). Such distortion of "information flows to
obtain personal benefit is a pervasive phenomenon... (causing)
increased costs of monitoring ... " (Cable and FitzRoy, 1980, p. 102).
Second, slow evaluation processes in the context of lifetime
employment also discourage deception since successful deception is
unlikely over long periods (Crawford, 1983, p. 9). Third, the open
physical space of the typical Japanese office where everyone can
notice everyone else contributes to low monitoring costs, especially

monitoring by peers (p. 9). Finally, peer monitoring is more likely to
occur in the presence of (1) team spirit and loyalty and (2) bonuses of
a profit-sharing nature (Cable and FitzRoy 1980, p. 103; Crawford,
1983, p. 7).23
In addition to higher motivation, one might expect improved
membership behavior (for example, lower absenteeism and turnover)
in Z organizations. Assume for the moment that the nature of a Z
firm's IPC makes it possible for workers to experience greater
current and future utility at given levels of current effort than A
workers, but no additional V' with extra effort. This is the situation
depicted in Figure 3; p. is the same for the Z and the A organization
even though V!] exceeds V';. Since the p' level is the same for both,
Z management has no motivation/effort advantage, but the Z worker
will receive relatively larger satisfactions by going to work each day
or staying with the company as opposed to either engaging in
nonwork activities or moving to another company given their
satisfactions. 24 According to this argument, the Z worker with a
morning headache or other good excuse for not going to work is less
likely to use the excuse. Since absenteeism and turnover impose real
costs on corporations, improved membership behavior may represent
a real productivity improvement. My view is that the satisfactions
from working in a Z organization lead to both higher worker effort
U'
U'*
Z
U'*
A
E* E
FIGURE 3
The present value of utility versus effort relationship.
218 JOHN F. TOMER
z
and improved membership behavior, i.e., curve U is above and to
the right of U~.
There is also reason to believe that organizational inertia will be
lower in Z organizations, or more generally, lower for organizations
further to the right on the IPC spectrum. The idea of inert areas
refers to situations where individuals or groups could reach higher
utility levels by adjusting their effort levels but find that it is not
worthwhile to do so (Leibenstein, 1976, pp. 111-114). Thus,
individuals in inert areas are within a range insensitive to external
changes that affect the utility they could realize. Leibenstein offers a
variety of reasons for inert areas including exaggerated assessments of
the temporary cost of moving, the fear of newness, fear of disap-
pointment, uncertainty, fear of conflict with others, and fear of
disapproval. Inert areas are important and, seemingly, inevitable
according to Leibenstein. My view is that inert areas are not
inevitable but to a great extent a product of workers' psychological
and economic insecurity. Therefore, when a company succeeds in
reducing workers' insecurity (such as would be expected when IPCs
move rightward) organizational inertia should be reduced. Reduction
in inert areas implies that workers' responsiveness to the
organization's demands will be increased. Thus, we would expect Z
workers to be quicker to adjust the quantity and quality of their
effort to changing organizational requirements, thereby increasing
organizational effectiveness.
4. CONCLUSION
Japan's superior productivity growth over the last two to three

decades is beyond question. But how much of this growth derives
from superior management? Indeed, what is it about Japanese
management that is superior and how does this translate into higher
productivity? My analysis, which builds on the analytical
frameworks of Oliver Williamson and Harvey Leibenstein, has
attempted to supply a numer of important arguments directed toward
answering these questions. In particular, Theory Z management, the
essence of the Japanese management ideal, reduces labor market
transaction costs, reduces the costs of noncooperation, increases
worker effort, reduces shirking costs, improves membership beha-

vior, and reduces inert areas in comparison to American (A)
management.
These benefits of Z management follow from the nature of
employer-employee relationships; in particular, Z relations are
characterized by high cohesiveness or solidarity. On the one hand,
the Z organizational incentives lead self-interested employees to
adopt behavior that is generally more favorable to the long-term best
interests of their organizations than in A organizations. On the other
hand, participants in Z organizations share much in the way of
common purposes and, ultimately, morality, thereby keeping self-
interest-seeking in check in such a way that disintegrative conflict
and other dysfunctional organizational behavior is discouraged (see
Macneil, 1980, especially pp. 95-97).
Much work, of course, remains to be done. For example, one
direction for future research would be to explore the extent to which
Z management provides incentives for rational decision making and
conflict resolution. Another direction would involve the empirical
analysis necessary to test the hypotheses posited in this paper. Edgar
Schein (1981) has raised some important questions regarding the
transferability of the Japanese management style to American busi-
nesses: (1) are the basic cultural premises in the U.S. too different
from Japanese ones to allow the acceptance and beneficial utilization
of Z management? and (2) what kinds of organization stand to
benefit most (least) from Z management? In commenting on an
earlier version of this paper, Jay Barney (1983) has suggested that Z
management may only be efficient for a certain class of transactions
between organizations and their workers, i.e., where employees have
made highly transaction specific investments, and may not be
efficient where employees' skills are general and could be utilized
equally well in other firms and types of transactions. Clearly, these
issues are deserving of further investigation.
Barney (1983) has also noted the strategic implications of the
thesis developed here. He points out that most firm strategies require
the acquisition of resources, but that the prices of these resources in
competitive markets practically eliminates the possibility of firms
using them to earn supernormal returns. A strategy for achieving
supernormal returns which may be open to firms, however, is that of
220 JOHN F. TOMER
better exploiting the resources already under their control. Thus, a

corporate strategy founded on developing Theory Z management
may very well be a more reliable path to supernormal returns than
those based on developing market imperfections or on acquiring an
optimal portfolio of assets. Paradoxically, the emphasis of Theory Z
is on people, not on profit.
It is important to note that the characteristics of Z-style implicit
psychological contracts make not only for high X-efficiency but for
humane work places. If, as it appears, Z employees have a greater
opportunity for self-actualization-realization of their full human
potential---one can say that Z firms are more a-efficient (McNally,
1981, p. 148) than A firms, and perhaps, that an economy dominated
by Z organizations is more a-efficient. From the standpoint of
psychological health alone, it seems that society would be better off
having more Z firms or at least more organizational investments in
moving IPCs rightward. 25
Will Z management become important in the u.s.? Because of
recent publicity regarding the virtues of Japanese management and
because of increased competition, especially from Japanese com-
panies, many U.S. firms are learning more about Japanese man-
agement and adopting certain of its features. My impression of these
changes thus far is that most of this imitation relates to particular
features or techniques-for example, quality circles-and not to the
whole of Z management. 26 That is, the organizational changes
involved represent at best small movements along the IPC spectrum.
It will be interesting to see whether bigger moves will be comingY
Another noteworthy trend is that more and more Japanese firms are
setting up shop here, apparently with considerable success, using a
version of Japanese management adapted to u.s. realities. 28 Finally,
"excellent" U.S. companies generally have at least some of the Z
management characteristics, especially the implicit control through a
strong corporate culture. 29
ACKNOWLEDGMENT
This chapter has benefited from the presentation of earlier vers-

ions to the Rensselaer Polytechnic Institute faculty, to a meeting of the
Association of Social Economics, at the Second U.S.-Japan Business

Conference, and at the Wharton Conference on Productivity, Tech-
nology, and Organizational Innovation sponsored by the Center For
The Study Of Organizational Innovation. Professors Benjamin
Gilad, Rutgers University at Newark, Jeremiah Sullivan and Thomas
Roehl, University of Washington, and Jay B. Barney, University of
California, Los Angeles, are noteworthy for their constructive
criticisms and comments on the chapter.
NOTES
1. Transaction costs refer to the costs of carrying out transactions In

markets including the costs of making a satisfactory contract.
2. Robert Hayes and William Abernathy (1980) have argued that the typical
characteristics of American management during the 1970s especially have
made it relatively inefficient.
3. Bureau of Labor Statistice, News, June 2, 1982.
4. COWPS, Report on Productivity, July 1979, as quoted in Hayes and
Abernathy (1980).
5. A study by Dale Jorgenson and Nieko Nishimizu (1978, p. 723) finds
similarly that by 1974 the gap between U.S. and Japanese technology had
been closed.
6. Insofar as Japanese management leads to lower worker resistance to
labor-saving innovations, Denison and Chung (1976, p. 82) suggest a yes
answer.
7. The Z management referred to in this article differs slightly from that of
Ouchi (see Ouchi and Price, 1978, p. 39). My version is closer to the
characteristics of the Japanese management ideal (see Ouchi, 1981, p. 58).
This difference reflects our different purposes.
8. The notion of Theory Z as the next higher level on the continuum
beginning with Theory Y was first developed by Abraham Maslow
(1971, pp. 270-286). Maslow's characterization of Theory Z organization
is "deeper" than Ouchi's in the sense that it is less a list of specific
organizational features and more of a characterization of the orientation
and behavior of individuals in a hypothetical Z organizational setting.
Such individuals are not only self-actualizers with high motivation and
commitment to task and team but are able to transcend the concrete
reality of organizational experience.
222 JOHN F. TOMER
9. In Japanese organizations, the guiding philosophy or spirit of harmony

(or wa) which derives from the Confucian cultural tradition has a very
important influence. See, for example, Riggs and Seo (1979).
10. The difference in the principal-agent relationship between Japanese and
American management is suggested by the different attitudes of Harold
Geneen, formerly chief executive officer of ITT, and Konosuke Mat-
sushita, founder of Matsushita Electric Company. According to Pascale
and Athos (1981, p. 83), "Geneen seemed to regard other people as
objects to be used to achieve his purpose, while Matsushita seemed to
regard them as both objects to be used and subjects to be honored in
achieving his and their purposes."
11. Ian Macneil (1980, pp. 10-35) uses a similar spectrum which starts with
discrete transactions and ends with the relational contract (especially the
modern variety as opposed to primitive ones). The IPC spectrum in
contrast relates to implicit contracts between employers and employees.
On the left of the IPC spectrum, the employment relation is very much
the same as Macneil's discrete transaction. Moving rightward along the
spectrum we first come to relations where the social cohesiveness among
employer and employee is relatively low, and thus, self-interested
maximizing behavior predominates. Moving further right we finally
come to relations where social solidarity is very high and where
self-interest has a much more circumscribed role to play. Ouchi (1980),
according to one interpretation (Tomer, 1983a, p. 60), views Theory Z or
clan style management as one of three distinct types of relations between
employers and employees rather than a spectrum of possible
relationships.
12. This hypothesis is intended to apply only in "modern" societies in which
economic organizations utilize systematic, rather than traditional, meth-
ods in striving for economic success.
13. Writing in the New York Times (March 30, 1983), John Holusha cites
one type of evidence of lower labor market transaction costs in Japan.
Holusha states that " ... labor contracts in Japan are models of brevity
compared with the bulky, legalistic documents that emerge from
negotiations in the United States. This freedom from encumbering detail
is said by analysts to be an important source of Japan's advantage over
American automakers."
14. Another possible pattern is nonreciprocal cooperation-at least non-
reciprocal among the parties involved. An example would be the
mentoring relationship in which the mentor helps a junior member of the
organization without the expectation of receiving help directly from the
junior employee. There may, however, be the implied expectation of

reciprocation to the organization as a whole. It is well known that
mentoring is an established part of Japanese management. While
mentoring occurs in American companies, the practice appears to be less
established and less encouraged.
15. Noncooperative behavior will be accentuated when organizations (1)
measure employee contributions narrowly and (2) promote high com-
petition among employees for organization rewards. Specialized career
paths may make communication difficult, and thus inhibit cooperation.
16. For a more complete analysis of cooperation, see Tomer (1983b).
17. A similar observation is made by Leibenstein (1982a, p. 9).
18. In a recent communication, Leibenstein (1982b) acknowledges that
individual performance may get worse (implying lower effort) as
pressure gets "too great." This, he indicates, is a departure from the
assumption typically implicit in his writings.
19. In slightly more detail, U,(E o) represents the total expected benefits (or
utility) arising from effort Eo expended in the current period, but
realized in period t. As an example, U,(E o) might be thought of as the
share of period t profits which a worker expects if his current effort is
Eo. Clearly, U,(E o) involves a complex assessment process, accounting
for the relationship between Eo (one's own current effort) and many
other variables, such as other workers' current and future effort,
management decisions, etc. We make no attempt to spell out how Ut(Eo)
is actually calculated, but simply use it here to make several points
regarding the impact of expectations of future benefits from current
effort decisions.
20. It is quite possible that workers in their effort decisions will consider
future satisfactions (from current effort) expected to be received as
members of a different organization. However, since the probability of
their realization is likely to be lower, these utility expectations are not
considered here.
21. This analysis is similar to psychologists' expectancy theory (see, for
example, Lawler and Rhode, 1976, pp. 19-22). In contrast to the present
analysis, expectancy theory emphasizes two steps, work efforts leading
to probable performance, which, in turn, leads to probable outcomes
more or less valued by workers. Workers are presumed to be rational in
the sense of attempting to maximize the value of desired outcomes in
their effort decision making.
22. Japanese workers may have low rates of time preference even before they
enter work organizations for reasons related to their culture, rearing, and
224 JOHN F. TOMER
school adjustment. However, these arguments are strictly unrelated to

the functioning of Z organizations.
23. FitzRoy and Mueller (1977, pp. 15-16) conclude that firms characterized
by homogeneity, small numbers, consensus decision making, high trust,
much informal communication, and cohesion will save monitoring costs
and minimize shirking.
24. Strictly speaking, the argument for reduced turnover would not apply
unless workers' employment alternatives do not have Z organizations. In
Japan, low turnover is encouraged by the diminished status and
opportunities of midcareer entrants. The latter seems largely to reflect
the time it takes to become part of the new organization's community
and internalize its values.
25. Tomer (1981 b) has pointed out that the psychological health benefits of
improved work environments are positive externalities or spillovers, and
therefore, firms taking only private incentives into account will tend to
underinvest in such organizational change. It seems plausible, however,
that if a substantial number of firms in a society were to move to the
right on the IPe spectrum, these external societal benefits would tend to
get internalized to the firm.
26. Essentially the same point is made by Peters and Waterman (1982, p.
241).
27. There are, of course, alternative strategies and courses of action which
organizations may choose to increase productivity through transforma-
tion of the IPe. The growing and voluminous literature on organiza-
tional development includes many examples.
28. This, of course, is supportive of the contention that the virtues of Z
management are ones that can be transferred successfully to non-
Japanese cultures. Pascale and Athos (1981) acknowledge that Japanese
culture provides advantages, but they do not believe that the features of
American culture represent an insuperable obstacle.
29. "Without exception, the dominance and coherence of culture proved to
be an essential quality of the excellent companies" (Peters and Water-
man, 1982, p. 75).
REFERENCES
Alchian, A. A., and Demsetz, H. (1972). Production, information costs and

economic organization, Am. Econ. Rev. 62, 777-795.
Argyris, C. (1960). Understanding Organizational Behavior, The Dorsey

Press, Homewood, Illinois. (1983).
Barney, J. B. (1983). Theory Z, Institutional Economics, and the Theory of
Strategy, Paper presented at Wharton Conference on Productivity,
Technology, and Organizational Innovation.
Byrd, R. E. (1974). A Guide to Personal Risk Taking, AMACOM, New
York.
Cable, J. R., and FitzRoy, F. R. (1980). Productive efficiency, incentives and
employee participation: Some preliminary results for West Germany,
Kyklos 33(1), 100-12I.
Clark, R. (1979). The Japanese Company, Yale University Press, New
Haven.
Coase, R. H. (1937). The nature of the firm, Economica (New Series), 4,
386-405.
Crawford, R. G. (1983). Organizational Structure and Productivity: The
Economics of Theory Z, Paper presented at the Second U.S.-Japan
Business Conference, Tokyo.
Denison, E. F., and Chung, W. K. (1976). How Japan's Economy Grew So
Fast: The Sources of Postwar Expansion, The Brookings Institution,
Washington, D.C.
FitzRoy, F. R., and Mueller, D. C. (1977). Contract and the Economics of
Organization, Paper presented at Fourth Interlaken Seminar on
Analysis and Ideology.
Gabarro, J. (1979). Socialization at the top-how CEOs and subordinates
evolve interpersonal contracts, Organ. Dyn. Winter, 3-23.
Grayson, C. J. (1982). Emphasizing capital investment is a mistake, Wall St.
j., October II.
Hayes, R. H., and Abernathy, W. J. (1980). Managing our way to economic
decline, Harv. Bus. Rev. 58, 67-77.
Jorgenson, D. W., and Nishimizu, M. (1978). U.S. and Japanese economic
growth, 1952-1974: an international comparison, The Econ. j. 88,
707-726.
Lawler, E. E., and Rhode, J. G. (1976). Information and Control in
Organizations, Goodyear Publishing Co. Santa Monica, California.
Leibenstein, H. (1976). Beyond Economic Man: A New Foundation for
Microeconomics, Harvard University Press, Cambridge, Massachusetts.
Leibenstein, H. (1982a). The Japanese Management System: An X-Efficiency
Game Theory Analysis, Harvard Institute of Econimic Research, Dis-
cussion Paper No. 938.
226 JOHN F. TOMER
Leibenstein, H. (1982). "Worker Motivation and X-Efficiency Theory: A

Comment," j. Econ. Issues 16, 872-873.
Levinson, H., Price, Charlton R., Munden, Kenneth j., Mandel, Harold j.,
Solley, Charles, M. (1963). Men, Management and Mental Health,
Harvard University Press, Cambridge, Massachusetts.
Likert, R. (1961). New Patterns of Management, McGraw-Hill, New York.
Macneil, I. R. (1980). The New Social Contract: An Inquiry into Modern
Contractual Relations, Yale University Press, New Haven.
Maital, S., and Maital, S. (1977). Time preference, delay of gratification and
the intergenerational transmission of economic inequality: A behavioral
theory of income distribution," in Essays in Labor Market Analysis,
(Ashenfelter, O. c., and Oates, W. eds.), Wiley, New York.
Marsh, R. M., and Mannari, H. (1976). Modernization of the japanese
Factory, Princeton University Press, Princeton, New Jersey.
Maslow, A. H. (1971). The Farther Reaches of Human Nature, Penguin
Books, New York.
McNally, M. (1981). On X-efficiency,]. Post Keynesian Econ. 4, 145-148.
Okun, A. M. (1981). Prices and Quantities: A Macroeconomic Analysis. The
Brookings Institution, Washington, D.C.
Ouchi, W. G. (1981). Theory Z: How American Business Can Meet the
japanese Challenge, Addison-Wesley, Reading Massachusetts.
Ouchi, W. G., and Price, R. L. (1978). Hierarchies, clans, and theory Z: A
new perspective on organizational development, Organ. Dyn. Autumn,
25-44.
Ouchi, W. G. (1980). Markets bureaucracies and clans, Admin. Sci. Q. 25,
129-14l.
Pascale, R. T., and Athos, A. G. (1981). The Art of japanese Management:
Applications for American Executives, Simon and Schuster, New York.
Peters, T. J., and Waterman, R. H. (1982). In Search of Excellence: Lessons
from America's Best-Run Companies, Harper and Row, New York.
Riggs, J. L., and Seo, K. K. (1979). Wa: Personnel factor of Japanese
productivity, Ind. Eng. April, 32-35.
Thomas, R. R. (1978). Managing the Psychological Contract, Harvard
Business School Case Services, Boston.
Schein, E. H. (1981). Does Japanese management style have a message for
American managers? Sloan Manage. Rev. 23, 55-68.
Tomer, j. F. (1981a). Organizational change, organizational capital and
economic growth, East. Econ.]. 7, 1-14.
Tomer, J. F. (1981b). Worker motivation: A neglected element in micro-
micro theory,j. Econ. Issues 15, 351-362.
Tomer, J. F. (1983a) The Japanese company: The challenge for economists,

East. Econ.]. 9, 57-68.
Tomer, J. F. (1983b). Productivity Through Cooperation: The Role of
Japanese and Gandhian Development Strategies, unpublished paper
prepared for the Third World Congress of Social Economics at Fresno,
California.
Vogel, E. (1979). Japan as Number One: Lessons for America, Harper and
Row, New York.
Wilkins, A. L., and Ouchi, W. G. (1983). Efficient cultures: Exploring the
relationship between culture and organizational performance, Admin.
Sci. Q. 28, 468-481.
Williamson, O. E. (1981). The modern corporation: Origins, evolution,
attributes,]. Econ. Lit. 19, 1537-1568.
Williamson, O. E. (1979). Transaction-cost economics: The governance of
contractual economics,]. Law Econ. 22, 233-261.
CHAPTER 8
THEORY Z, INSTITUTIONAL
ECONOMICS, AND THE
THEORY OF STRATEGY
Jay B. Barney
1. INTRODUCTION
Tomer's (1983 and Chapter 7, this volume) discussion of the

microeconomic foundations of Ouchi's (1981) Type Z management is
provocative not only because it begins to explore the underdeveloped
relationship between microeconomics and management theory, but
also because it discusses both of these theoretical traditions in an
explicitly strategic context. In this response to Tomer, I will first
continue to explore the relationship between management theory and
microeconomics. After these specific comments, I will then turn to
the more general issues raised by Tomer, i.e., the use of the theory of
management and microeconomics to develop a theory of business
strategy.
2. MICROECONOMICS AND TYPE-Z MANAGEMENT
Tomer's (1983) basic thesis concerning the relationship between

microeconomic theory and Type-Z management can be briefly
summarized. The description of management systems whose
/try B. Barney. Graduate School of Management, University of California, Los

Angeles, California 90024.
229
230 JAY B. BARNEY
effectiveness rests on the development of long-term, qualitative

relationships between economic factors, far from being inconsistent
with economic theory, can in fact be derived from economic theory.
However, the economics which should be used in this derivation
process is not neoclassical in nature, but rather some form of the new
institutional economics. Tomer (1983 and currently in Chapter 7)
discusses, in particular, Williamson's (1975, 1979) transaction costs
economics and Leiberstein's (1976) theory of X-efficiency to suggest
that Type-Z managed firms are, in fact, reducing the transaction costs
associated with the employment relation. The primary mechanism for
accomplishing this reduction in transaction costs, according to
Tomer, is what he calls an implicit psychological contract, what
Gabarro (1979) has called an interpersonal contract, and Williamson
(1979) has called a bilateral contract.
This basic thesis is hard to dispute, nor will we do so here.
Indeed, Ouchi's original Theory Z research has itself developed along
the lines suggested by Tomer. In 1980, Ouchi published a paper
titled "Markets, Bureaucracies, and Clans," which provides an
explicit transactions costs rationale for the development of implicit
interpersonal contracts. This was followed by two additional pieces
which have further refined the micro economic reasoning behind
Type-Z management, one titled "Efficient Cultures" (coauthored
with Alan Wilkins), and a second titled "Information Cost and the
Organization of Transaction Governance" (coauthored with myself).
Moreover, other management theorists, including Andy Van de Ven
from the University of Minnesota, are using various transaction costs
arguments to underpin their arguments.
Despite the basic soundness of Tomer's observations, and the
growing literature consistent with it, care must be taken when
comparing and contrasting Theory Z, as a theory of management,
and institutional economics. The need for care is particularly
important when deriving normative implications about how to
manage a firm. In particular, confusion with reference to the
theoretical unit of analysis can lead, in such normative work, to an
overreliance on nonmarket forms of governing the employment
relation.
Much of the strength of the transaction costs framework
developed by Williamson (1975) is its emphasis on the transaction as
THEORY Z, INSTITUTIONAL ECONOMICS 231
the unit of analysis. The basic logic here begins by observing that
transactions have different characteristics and attributes, and that as
these characteristics vary, different means of governing the relation-
ship become more or less efficient. Following Williamson (1979) and
Barney and Ouchi (1983), the more transaction specific the in-
vestments required to carry a transaction forward, the more elaborate
are the nonmarket forms of governance required to manage the
transaction.! If transaction investments are not specific, then market
forms of transaction governance are efficient, and thus more desir-
able, than nonmarket forms.
The implication of this logic for the integration of Theory-Z
management and institutional economics is clear. Theory Z describes
the mechanisms for governing the employment relation under
conditions of high transaction specific investment. Under conditions
of low transaction specific investment, Theory Z, with its reliance on
nonmarket implicit psychological contracts, will be inefficient. Thus,
in applying institutional economics to the analysis of the employment
relation, a transaction-by-transaction approach must be taken. For
those employment relations characterized by high transaction specific
investments, Type-Z management should be acopted. For
employment relations without transaction specific investments,
Type-Z management should not be adopted.
One is, of course, forced to ask the question "Do employment
relations with low transaction specific investments exist?" The
answer must certainly be in the affirmative. Consider the simple
example of a long-haul moving company. In moving clients from one
part of the country to another, at least three types of employees are
relevant, and each has different levels of transaction specific in-
vestments. The first type of employee packs the customer's goods in
boxes and then loads the truck. The second type of employee drives
the truck from customer to customer, supervises the packing, and
loading, and once the truck is full, drives the truck to its destination
and supervises the unloading/unpacking. The third type of employee
manages this process, fills out required FTC forms, obtains in-
surance, advertises for customers, etc. The first type of employee has
basically no transaction specific investments: large numbers of in-
dividuals could each do this essentially standard task and no special
training beyond a few essentials concerning packing and loading is
232 JAY B. BARNEY
needed. Because there are no transaction specific investments, this

relationship can be managed most efficiently through market mecha-
nisms. And, indeed, in this industry, packers and loaders are usually
hired out of union halls on a spot market, standard short-term
contract basis. This differs substantially from the truck driver who
may have to develop specialized skills in dealing with government
regulation in the moving industry, and who also interfaces most
directly with the customer. Moreover, the truck driver, because he
supervises packing and loading, bears significant, nontransferable
responsibility for the safety of the load. Because of these and other
transaction specific investments, spot market contracts will not be
adequate to govern this relationship, and will be replaced by more
elaborate, long-term contracts. However, transaction specific in-
vestments are not so high so as to suggest that the truck driver should
become a regular full-time employee of the moving firm. Instead, the
truck driver is normally an independent contractor with a long-term
contract with the moving company. Finally, for managers who
develop an enormous reservoir of information concerning how the
moving business works, how it is sold to customers, how to deal
with government regulations, how to manage relations with in-
dependent truck drivers, etc., transaction spcific investments are very
high, and the most efficient way to govern this employee relation is
for the manager to become a full-time employee of the firm.
Now, under which of these conditions is it appropriate for
Theory Z concepts to be applied in managing the employment
relation? Clearly, Theory Z concepts are not applicable in the packer
and loader case. The use of implicit psychological contracts under
these conditions would be unnecessary, for no long-term relations
with high interpersonal content are contemplated. Just as clearly,
Theory Z concepts are applicable in the management case. Here, high
transaction specific investments require the type of governance
mechanisms which allow for long-term stable transactions. Between
these two extremes we find the truck drivers, who will, because of
the long-term contract they sign, develop some implicit psychological
contracts with the firm. Yet these implicit contracts are in addition to
explicit long-term contracts which form the primary basis of the
relationship. Thus some Type Z management mechanisms may be
appropriate in this case. The key point, again, is that Type-Z

management must be applied on a transaction-by-transaction basis.
The efficiency of Type-Z management ultimately depends on the
characteristics of the employment relationship under consideration.
Within this context, transaction costs economics can be a very
valuable tool in the development of a contingency theory of
management, and Theory Z represents one important class of
management tools that is appropriate for governing transactions
under certain specifiable conditions. Notice that this contingency
application of Type-Z management is not inconsistent with Ouchi's
(1981) original work. Ouchi studied the employment relations of
workers and management in large organizations in the U.S. and in
Japan. It is precisely these types of relations that are probably most
frequently characterized by high transaction specific investments,
where the implicit psychological contracts of Theory Z are most
appropriate.
3. INSTITUTIONAL ECONOMICS, MANAGEMENT, AND

THE THEORY OF STRATEGY
At a more general level, Tomer's paper, especially because it is

presented at a conference that includes sessions on manufacturing
strategy, raises the question of the strategic value of management
theory and institutional economics. Historically, the study of
strategy has not had a strong discipline base. However, with work by
Caves and Porter (1977) and Porter (1980), to name just two
important pieces, economics has begun to emerge as the underlying
paradigm for strategy. However, the economics used by these
authors in their analyses is neoclassical in nature. In particular, Porter
(1980) outlines the underlying market characteristics that in the
neoclassical model lead to normal returns. He then suggests that
firms, in order to obtain supernormal returns, should attempt to find
industries that are not characterized by the normal return generating
attributes, or, if this is not possible, firms should attempt to create
these competitive imperfections themselves. In this way, strategizing
can lead to firms obtaining supernormal returns.
234 JAY B. BARNEY
Though this approach to understanding strategy has made great

strides, it may ultimately be very limiting. Indeed, it may be the case
that institutional economics, as opposed to neoclassical micro-
economics, will provide a more fruitful paradigm base for the theory
of strategy. And because, as Tomer and others have noted, the theory
of management is wholly consistent with this institutional economics,
we can equivalently suggest that the theory of management may be a
more fruitful paradigm base for developing a theory of strategy than
neoclassical economics. The potential that institutional economics
and management theory hold for developing a theory of strategy is
discussed in detail elsewhere (Barney, 1983a, 1983b). We will only
review the major points of the argument here. We begin by
observing, along with the neoclassical microeconomists, that super-
normal returns can only be obtained by firms facing imperfectly
competitive conditions. However, though the existence of imperfect
competition may generate supernormal returns, this does not neces-
sarily imply that the creation of conditions of imperfect competition
will generate supernormal returns. The creation by a firm of
imperfectly competitive conditions is a manifestation of that firm
implementing a strategy. The strategy might be putting barriers to
entry in place, it might be negotiating favorable positions vis-a-vis
supplies, or it might be negotiating favorable positions vis-a-vis
buyers (Porter, 1980). The implementation of such strategies requires
resources. For many firms, it may be necessary to acquire strategy
relevant resources. For example, one barrier to entry may be
low-cost production. According to Henderson (1979), low-cost
production will necessarily imply high-volume production, which in
turn implies the need for large market share. Thus if a firm wishes to
implement a low-cost producer barrier to entry, following this logic,
the resource it may need to acquire is market share. If a firm wishes
to implement a diversification strategy, through acquisition, the
resources it may need to acquire are the technology and business of
another firm.
If, in order to implement strategies that create imperfectly
competitive markets, resources must be acquired, then the possibility
of obtaining supernormal returns will depend not only on whether or
not a market imperfection is created by the strategy, but also upon
the price one pays for the resources acquired. It can be shown that,
under a wide set of circumstances, these resources will be ap-
propriately priced when acquired, i.e., they will have a normal return
associated with their acquisition. A simple example may help clarify.
Suppose a firm wishes to acquire a second firm, and in so doing,
create an imperfectly competitive market and obtain supernormal
returns. This acqusition takes place in the market for companies.
Research has shown (Copeland and Weston, 1979) that this market is
quite competitive in the sense that acquiring firms can usually expect
to obtain only normal returns from acquisitions. Thus, the acquisition
designed to create imperfect competition is appropriately priced, and
the acquiring firm obtains only normal returns. This form of analysis
can be applied to any resource a firm must acquire to implement a
strategy, whether it is market share to obtain learning curve
advantages (Henderson, 1979), high investment in manufacturing
technology to create a barrier to entry (Caves and Porter, 1977), or a
long-term relationship with a supplier (Porter, 1980). If a competitive
market for the resources can develop, these resource acquisitions will
only lead to expected normal returns. 2
How is it, then, that firms can obtain expected supernormal
returns. The answer must be that a firm should attempt to implement
strategies that exploit strategy relevant resources it already controls.
That is, it must avoid implementing strategies that require it to
acquire strategy relevant resources. These resources could have been
acquired for one purpose but have strategic value in another
application, or they could be part of the initial resource endowment
of the firm. In either case, strategizing for expected supernormal
returns reduces to the analysis of assets currently controlled by a firm
together with the exploitation of these assets in developing and
implementing strategies. Because these assets and resources do not
have to be acquired, they may be able to generate expected
supernormal returns. Institutional economics and the theory of
management become relevant in strategizing then, because these
theories can be applied by managers in isolating the key resources
they control. Once these theories point to what the key resources
are, guidance can then be given with reference to how they can be
strategically exploited. That is, because institutional economics and
236 JAY B. BARNEY
the theory of management can help firms focus inwardly on the

resources they already control, these theoretical traditions have
relevance in firm strategizing for expected supernormal returns.
In this brief summary of the argument, numerious subtitles have
been left out. For example, luck continues to playa role even when
applying institutional economics and the theory of management in
isolating strategy relevant resources already controlled by a firm.
Nevertheless, this approach to strategizing seems to hold significant
potential, a potential that Tomer begins to address when he considers
the relationship between Theory Z and institutional economics.
Nevertheless, much of this potential is unrealized, for institutional
economists and management theorists have only recently begun
exploring the strategic implications of their research. For me, one of
the great potentials of a book like this is to further these
developments.
NOTES
1. A transaction specific investment is some investment, either of the

physical or human capital type, that has maximum value in one particular
transaction, and whose value in any other transactions is considerably
less. In other words, transaction specific investments are investments that
cannot be applied to other than the current transaction without
significant reduction in the value of the investment.
2. It may, of course, be possible for a strategy relevant resource to have an
uncertain value in implementing a strategy. Under conditions of un-
certainty, firms may be able to acquire valuable resources at a price that
can lead to supernormal returns. However, because these returns are
attributable to uncertainty, they are unexpected supernormal returns, a
manifestation of a firm's good fortune and luck rather than its
strategizing skill.
REFERENCES
Barney,]. B. (1983a). Strategy and Organization, University of California at

Los Angeles Report.
Barney, J. B. (1983b). Rational Expectations, Markets for Strategy

Implementation: Asymmetric Expectations, Luck, and the Theory of
Strategy, University of California at Los Angeles Report.
Barney, J. B., and Ouchi, W. G. (1983). Information Cost and the
Organization of Transaction Governance, University of California at
Los Angeles Report.
Caves, R., and Porter, M. (1977). From entry barriers to mobility barriers:
Conjectural decision and contrived deterence to new competition, Q. j.
Econ., 91,241-281.
Copeland, T., and Weston, J. F. (1979). Financial Theory and Corporate
Policy, Addison-Wesley, Reading, Massachusetts.
Gabarro, J. (1979). Socialization at the top-How CEO's and subordinates
evolve interpersonal contacts, Organ. Dyn., 7(3), 3-23.
Henderson, B., (1979). Henderson on Corporate Strategy, Mentor, New
York.
Lieberstein, H. (1976). Beyond Economic Man: A New Foundation for
Microeconomics, Harvard, Cambridge, Masachusetts.
Ouchi, W. G. (1981). Theory Z: How American Business Can Meet the
japanese Challenge, Addison-Wesley, Reading, Massachusetts.
Porter, M. (1980). Competitive Strategy, Free Press, New York.
Tomer, J. (1983). Working Smarter the japanese Way: The X-Efficiency of
Theory Z Management, Union College Report.
Wilkins, A., and Ouchi, W. G. (1983). Efficient cultures, Admin. Sci. Q. 28,
468-481.
Williamson, O. (1975). Markets and Hierarchies: Analysis and Antitrust
Implications, Free Press, New York.
Williamson, O. (1979). Transaction costs economics: The governance of
contractual relations,j. Law Econ. 22, 233-261.
CHAPTER 9
PRODUCTNITY
MEASUREMENT AND
MANAGEMENT CONTROL
Rajiv D. Banker
1. INTRODUCTION
In recent years, manufacturing firms in the United States have

been directing increasing attention to the need for improving
productivity in order to strengthen their competitive position. To
monitor and evaluate productivity changes, and to communicate and
reinforce this new emphasis on productivity improvements, ap-
propriate measures of productivity are required. Productivity mea-
surement can generate information that may be useful for man-
agement control purposes in two ways. First, such measurement may
identify dimensions on which productivity improvements are pos-
sible, in which case targets may be generated to guide future
operations. The productivity measures may be employed to identify
products or processes for which the firm has a competitive edge in
productivity or similarly evaluate new technologies or investment
projects that are likely to increase the firm's profitability. The second
role for productivity measures in management control systems arises
from their use in evaluating the performance of managers. In
conjunction with the usual profit or cost accounting system, a
productivity measurement system can be designed by the corporate
Rajiv D. Banker. School of Urban and Public Affairs, Carnegie-Mellon University,

Pittsburgh, Pennsylvania 15213.
239
240 RAJIV D. BANKER
management so that it motivates the divisional managers to select

actions that are consistent with the overall corporate plans and
priorities.
Recent surveys, however, have documented that most large
corporations do not use direct measures of productivity. Instead,
they continue to use surrogates for productivity, such as labor
variance, capacity utilization, or production cost standards.
Sumanth's (1981) survey of 61 manufacturing companies indicated
that partial productivity measures were used in 19.3 % of the cases,
aggregate productivity measures were used in only 0.3% of the cases,
and simple surrogate measures were used in 80.4% of the cases.
Furthermore, considerable confusion prevails about the definition
and implementation of productivity measures. This is likely due to
the fact that the direct productivity measures that have traditionally
been used are simplistic, and they generally do not reflect the
complex nature of most production processes encountered in prac-
tice. In particular, the traditional approaches assume linearity (or
constant returns to scale), separability (or independence of input
consumption requirements), and other restrictive features for the
production technology to simplify the calculations. As a result,
improvements in productivity as measured by the traditional ap-
proaches are frequently found, in practice, to be inconsistent with
improvements in the firm's competitive position and its short-term or
long-term profitability. This arithmetical simplicity may have been
necessitated in the past by computational concerns. However, with
the increasing availability of inexpensive computing facilities, it is
imperative that new productivity measures be developed that are
appropriate for today's technologies.
A management control system is sometimes described as con-
sisting of (a) a device for measuring what is actually happening, (b) a
device for assessing the significance of the measured quantity relative
to a standard, and (c) a device for influencing the behavior of
managers, with evaluation of performance based on this measure and
standard.':' Similarly, a productivity measurement system requires
first a proper definition and development of a productivity index that
". See Anthony and Dearden (1980).

PRODUCTIVITY MEASUREMENT AND MANAGEMENT CONTROL 241
reflects the actual productivity considerations in complex technolo-

gies. Second, it requires benchmarks for comparison purposes. These
benchmarks, or standards, may be based on the information about
the organization's performance in the past. Such information need
not be restricted to just the performance in the preceding period. The
standards could also be based on the performance of other similar
organizations. This is particularly appropriate for corporations with a
number of similar plants or branches or divisions which are evaluated
relative to each other. Such benchmarks could also be developed
from extensive engineering simulation models. This approach is
appropriate for evaluating the productivity impact of new or
potential technologies.
Finally, a productivity measurement system must be integrated
with other systems for evaluating managerial performance. Produc-
tivity measurement is not very meaningful without an appropriate
accounting and reporting system to support it. However, currently in
most corporations, much of productivity oriented information re-
mains at the shop floor level and is rarely reported to higher levels of
management. If productivity accounting is to be employed for
management control purposes it must be defined and developed
carefully, and integrated with existing accounting and reporting
systems, and further, it must maintain consistency with other
corporate objectives such as profit maximization.
In this chapter, I shall describe a new framework for measuring
productivity, and for identifying areas of strengths and weaknesses
on this productivity dimension. By ensuring congruence between
productivity improvements and profitability enhancements, and by
generating targets and referent guides for productivity improvements,
this framework provides an appropriate basis for a management
control system. In addition, this approach also generates information
that is useful for analyzing decisions about investment in new
production technologies.
Productivity, as defined in this framework, measures the extent
to which the actual input consumption exceeds the minimum inputs
necessary to produce the actual output levels. The minimum input
consumption is determined in a manner consistent with our current
knowledge of the available production technology. Thus, a distinc-
242 RAJIV D. BANKER
tion is made between resources controllable by the responsible

manager, and resources that are fixed, or beyond his discretionary
control. In a manner analogous to responsibility accounting, the
productivity variations are analyzed to emphasize which manager
controls what parameters of the firm's operations, and to what extent
each manager is responsible for the observed productivity variations.
I shall begin by describing how the use of the traditional
productivity measures may lead to decisions that are inconsistent
with profit maximization or other corporate priorities. I shall next
describe the framework for productivity measurement and the
computational and data requirements for this approach. Finally, I
shall briefly discuss how this productivity analysis is integrated with
the usual management accounting practices of profit variance
analysis.
2. CONGRUENCE WITH PROFIT MAXIMIZATION
The traditionality used direct measures of productivity can be

categorized into three groups. The first consists of partial produc-
tivity measures such as materials usage per unit output, or value
added per labor hour, which has been adopted by the Bureau of
Labor Statistics. These are based on simple ratios of output quantity
divided by a single input level. The second group of measures
recognizes the multiple input nature of most production processes,
and therefore, employs a profile (vector) of partial productivity
measures. See, for instance, Gold (1980). The third approach,
described almost 30 years ago by Hiram Davis (1955), derives a single
aggregated productivity measure as the weighted sum of the partial
productivities for different inputs, where the weights are based on the
cost shares of individual inputs. However, all these approaches
impose strong assumptions of linearity (constant returns to scale),
and separability (independence of input requirements) on the un-
derlying production technology. Therefore, they frequently lead to
managerial decisions that are not consistent with profit maximization.
I shall illustrate this for a simple output: input partial productivity
measure, but the principles discussed here extend also to other
RELATIVE
OUTPUT
(UNITS) ..... .......... PRICES
M~----------~~-,
C' INPUT
(LABOR HOURS)
C = ACTIVITY LEVEL MAXIMIZING PROFITS

D = ACTIVITY LEVEL MAXIMIZING TRADITIONAL
PRODUCTIVITY MEASURE
FIGURE 1
Maximization of average productivity or profits (competitive benefits).
traditional productivity measures. These considerations will also be

employed to motivate a new approach to productivity measurement.
Consider an observation such as point B in Figure 1. The
average productivity of B is given by
output units OM B'B

~--~--~---=--- =---
input labor hours OB' OB'
But on considering the traditional productivity measure for the point

D, it is apparent that the productivity at D = D'D/OD' is greater
than B' B / a B', the productivity at B. Therefore, the productivity
measure can be increased simply by reducing the activity level to that
corresponding to D. Here, a reduction in output seems to improve
productivity.
For the price of the output relative to the labor usage rate
244 RAjIV D. BANKER
depicted by the relative price line in Figure 1, it is also apparent that

profits will be maximized if, in fact, the activity level is increased to
that corresponding to the point C. But this implies a reduction in the
traditional productivity measure since C' C / OC' is less than B' B /
OB'. In this situation, therefore, the criterion of increasing "produc-
tivity" leads to a decision that is contradictory to the objective of
increasing profits. This remains the fundamental problem with the
traditional methods, where productivity improvement does not
always imply profitability improvement. In fact, many corporations
experimenting with productivity measures have found that some
divisions making great strides in improving productivity have ac-
tually been showing lower profits. This is one of the principal
reasons why we have not seen the acceptance of productivity
measures as an integral part of a management control system, even
though considerable attention has been focused for so long on the
problem of productivity management.
The simple situation depicted in Figure 1 involves only output
measured in number of units. But this notion can be directly
extended to a wider range of competitive issues. For instance, a
reduction in labor hours may imply a lower quality of service or
product, even if the same number of units are still produced. Further,
such reduction may also affect the resilience of the organization to
respond to change, or the investment in terms of time and effort in
new products to improve the competitive position in the future. The
use of just the traditional productivity measures may, in this manner,
bias decisions against considerations for maximization of profits or
competitive benefits.
It is thus important to emphasize the impact of reduction in an
input, such as labor, on the level of output measured broadly,
possibly on multiple dimensions. Reduction in labor (or other inputs)
without reducing the output (on any dimension) is desirable-it is
true trimming of excess fat. But on the other hand, a decision to
reduce input consumption that will also reduce the output (on some
competitive dimensions) must be evaluated with due regard to overall
competitive considerations.
This discussion suggests an alternative approach to measurement
of productivity. Consider a point such as A in Figure 1. By
comparing A to the point B, it is again apparent that the level of

input consumption can be reduced from OA ' to OB' without
reducing the output level. The excess consumption B A reflects the
I I
excess fat that can be trimmed off. A productivity measure OB' /

OA I, keeping the output level constant at OM, monitors this
possibility for reducing labor hours. In the next section, I shall
provide a general framework for such productivity analysis embed-
ded in the usual production cost variance analysis.
3. A FRAMEWORK FOR PRODUCTIVITY ANALYSIS
The productivity measurement approach described in this sec-

tion builds on the pioneering work of Farrell (1957) in measuring
relative technical and allocative inefficiencies, and is based on the
theoretical constructs of production economics developed by
Shephard (1953). Farrell's early work was cast in a ratio form, with
an equivalent linear programming formulation, by Charnes, Cooper,
and Rhodes (1978). Its formal links with production economics were
axiomatized by Banker (1980) and Banker, Charnes, and Cooper
(1984), to extend the efficiency analysis to the case of nonconstant
returns to scale and distinguish between the purely technical and
scale effects. Banker (1984) introduced the notion of the most
productive scale size, and in (1983) suggested a framework for
productivity measurement and variance analysis. Banker and Main-
diratta (1983a) provide the link between this efficiency and produc-
tivity analysis and the neoclassical nonparametric production analysis
of Varian (1983).
For concreteness, I shall describe the productivity measurement
method for the case of a firm having two outputs and four inputs.
The actual identification of the principal outputs and inputs requires
the exercise of managerial judgment and will depend on the particular
circumstances for each firm. The estimation of productivity and other
indexes is based on the following data:
a. Actual values of inputs consumption, outputs produced, and

input and output prices (if known).
246 RAJIV D. BANKER
b. Similar values for a set of referent points relative to which

productivity attainment will be assessed. These referent
points will depend on the actual context of the application of
this model, but it could consist of time series data for the past
performance of the firm, cross-sectional data available for the
operating performance of other firms, or technical estimates
of various production possibilities in case a new technology is
being considered.
c. Additional knowledge about the actual production technol-
ogy, such as separability, substitutability or linearity of the
production function, and the controllable or fixed nature of
the input.
I shall consider the two products to be separable-i.e., the input

requirements for one product are not dependent on the level of
production of the other product. The productivity measured can then
be determined separately for each product, and in the developments
that follow, I shall refer to only Product Yl' If the process is not
separable, then the productivity measures will be jointly determined.
The analysis will be similar to that described here for the case of two
substitutable inputs. In the models that follow, Input 1 is labor
hours,':- considered to be controllable and having a separable impact
on the process. Input 2 is plant capacity, which is considered to be
fixed or noncontrollable in the short run, but other things remaining
the same, a lower capacity will cause congestion and greater materials
consumption. Inputs 3 and 4 are two types of materials which are
considered to be substitutable for each other.
I shall denote the n referent observations by the index j =
1, ... , n, and the actual observation whose productivity is being
evaluated will be indexed A. The output level for product Yl will be
denoted by YIA and Ylj, j = 1, ... , n. The input levels will be
denoted by XiA and X,j' where i = 1,2,3,4 and j = 1, ... , n. The
prices of the four inputs will be referred to as Ci, i = 1,2,3,4, and
assumed to be beyond the control of the manager.
I shall consider first the evaluation of productivity for product
"See Banker and Morey (1984) for a treatment of categorical variables.

Yl and input 1, labor hours, which is assumed to have a separable

impact on the process. The labor productivity Dl is mreasured by the
following linear program:
Dl = minhA (1)
subject to
n
2:
j~l
ajYlj :::::: Y1A
2:
j~l
ajXlj ::; hA x l A
The productivity measure defined above can be understood

better by referring to Figure 2. Consider first the point A °j its
productivity measure will be equal to MO BO I MO A 0 , which is less than
one. This indicates that there is a referent point BO, which produces
°
the same level of output as A with a smaller amount of the input.
The linear program in (1) will detect this possibility, and identify the
referent point BO, possibly as a convex combination of actual referent
points j = 1, ... , n . Targets for input reduction can then be set with
reference to the configuration of BO. On the other hand, consider the
point A':·j its productivity measure will be equal to M':-B':-I M':A\
which is greater than one. This implies that none of the referent
points, or their convex combinations, indicate that this level of
output can be produced with a level of input consumption equal to
or less than that measured for the observation A':-. Thus, we can infer
that this observation A ,:- reflects recent gains in productivity.
Next, I shall consider the measurement of productivity for the
two substitutable inputs 3 and 4. The combined materials produc-
tivity is measured by the following linear program:
(2)
248 RAJIV D. BANKER
OUTPUT
(UNITS)
o INPUT
(LASOR HOURS)
PRODUCTIVITY OF A * -
M* S*
M* A*
>, IMPROVEMENT
PRODUCTIVITY OF AO =
M'I
MO AO
SO
<, DEFICIENCY
FIGURE 2
Detection of productivity improvement.
subject to
n
L aYl' 2: YIA
j=l J J
Xi = L ajXij ::s hAXiA, i = 3,4

j=l
n
L
j=l
aX2'
J J
::s X2A
La
j=l J
= 1
Xi' hA , aj 2: 0
Note that no productivity measures are computed corresponding

to input 2, since plant capacity is considered to be fixed, and hence,
any reduction targets for this input are managerially irrelevant. This
fixed nature of plant capacity, and its relationship with materials
consumption, however, is modeled by ensuring that the constructed
referent points have the same or more constrained capacity levels.
This is reflected in constraints such as I:7=1 ajXZj :s: XZA in linear
program (2).
The productivity analysis in (2) for the situation of substitutable
inputs can be understood better by referring to Figure 3. The
diagram displays an isoquant for material 3 and material 4 required to
produce a given level of output YIA- For the relative prices of the two
materials, given by the slope of the broken line, it is evident that the
minimum costs will obtain at the point C. Geometrically, the ratio
OEIOA measures the combined productivity Di,4 defined in linear
MATERIAL 4
K~------~------~
............
............
RELATIVE PRICES
o MATERIAL 3
COMBINED PRODUCTIVITY = OE /OA

TECHNICAL PRODUCTIVITY = OB /OA
ALLOCATIVE PRODUCTIVITY = OE /OB
FIGURE 3
Productivity measurement with substitutable inputs.
250 RAJIV D. BANKER
program (2) as the ratio of the minimum cost (for the two materials)
to the actual cost.
It is further seen in this diagram that the point B, having the
same mix of the two inputs as the point A, can produce the given
level of output Y lA with smaller levels of consumption for both
materials 3 and 4. This suggests the disaggregation of the combined
productivity measure for materials into a purely technical produc-
tivity measure (n,4) and an allocative or mix productivity measure
(M~,4)' The first identifies productivity gains or declines because of
improvements in the production technology or process. The second
measures the productivity impact of the materials mix choice to the
extent appropriate substitutions between the two materials as dictated
by their relative prices were achieved. The purely technical produc-
tivity component is measured in Figure 3 as the ratio OBIOA, which
is less than one, and indicates a decline in technical productivity. The
allocative or mix productivity is simply the combined productivity
(D~,4) divided by the technical productivity (n,4)' and is measured
by the ratio OEIOB, which is less than one. This indicates that the
appropriate cost-minimizing mix of inputs (relative to the referent
possibilities) was not employed.
The numerical values for n,4 and M~,4 are obtained from the
following relationships:
(3)
(4)
where
x~ = g~X3A - S;A
x~ = g~X4A - S~A
and g~, S~A' S~A are optimal values of the variables in the following
programming formulation:
(5)
subject to
2:
j=l
ajY1j ;::: Y1A
2:
J=l
ajXij = gAXiA + 5iA, i = 3,4
n
2:
J=l
ajx 2j :s X 2A
2: a = 1
j=l J
gA' aj' 5iA ;::: 0
where f > 0 is a small non-Archimedean quantity.

The overall productivity measure for output Y1 for the point A
can now be defined in a manner that is a generalization of the Davis
(1955) approach. Thus, the overall measure is given by
(6)
where
It is easy to see that this overall productivity measure for output

Y 1 is the ratio of the minimum cost (based only on the referent
observations) to the actual cost of production for producing the
output level Y 1A' The overall productivity measure for the second
output Y2 can be computed and analyzed into its different com-
ponents in an analogous manner.
I shall turn next to the integration of this productivity measure-
ment method with profit variance analysis. The variance of actual
profits from the maximum possible profit~ can be analyzed into two
components. The first is the variance that arises because the actual
output level was different from the level that would have maximized
profits given the actually prevalent demand conditions for the
252 RAJIV D. BANKER
products. However, it is generally very difficult to assess the actual

demand conditions, the possible prices at which different levels of
output could be sold. See Banker (1983) for a discussion for how this
first variance may be computed, and further analyzed, if it is possible
to assess the demand function. Therefore, I shall focus attention on
the second component of the profit variance, due to the actual
variable production costs at the actual output levels exceeding the
minimum possible costs to produce this level of output. Thus,
Total variable production cost variance

= Variance for output y 1 production costs
+ Variance for output Y2 production costs
= (D 1 - 1) (Actual production costs for Y 1)
+ (D2 - 1) (Actual production costs for Y2) (7)
These variances can be further analyzed into the vanous

components of D 1 and D2 discussed above.
The methodology described above can also provide useful
information for assessing a new technology. The distance between
the existing and the projected production frontiers can be determined
using this model, and the expected productivity gain due to the
introduction of the new technology can be computed. This can be
illustrated with reference to Figure 4. Here x(y) denotes the
minimum amount of the input required to produce an output level Y
under the old technology, and z(y) denotes a similar quantity
projected under the new technology, where x and z are determined
for each Y using formulations as in (1). The probability of producing
a level of output is denoted by a density function p(y), this
probability function being derived from optimal output levels
corresponding to various prices that may prevail for the output. The
expected productivity gain is then measured by the expression
J:'x(y)p(y) dy / f2 Z (y)p(y) dy (8)
where [y l' y 2] represents the range of output levels considered

OUTPUT
(UNITS)
OUTPUT V t-----#
RANGE
o Zl Xl Z2 INPUT
Z(V) X(V) (LABOR HOURS)
f -:-f
V2 V2
PRODUCTIVITV GAIN = X(V) • PlY) dV Z(V)· PlY) dV
Vl Vl
FIGURE 4
Analysis for investment to improve productivity.
probable. This analysis can, thus, assist in the evaluation of the

imp cat of the new technologies on productivity.
4. CONCLUDING REMARKS
The framework for productivity analysis presented in this paper

provides a basis for incorporating productivity measures in a
management control system. These productivity and other indexes
are measured using a computationally efficient iterative linear pro-
gramming formulation, that can be implemented in most computer
environments. The estimation method is based on a very flexible
nonparametric method for production frontier estimation, and there-
fore, it is suitable for a wide range of production technologies
encountered in practice. On the other hand, traditional accounting
254 RAJIV D. BANKER
methods for the purpose tend to impose considerable structure on

the production technology, in order to simplify the arithmetic for
productivity measurement. This has made their applicability to many
real production situations highly limited. With the increasing availa-
bility of relatively inexpensive and powerful computational capabi-
lity, it has become imperative to explore generalized approaches that
will identify areas of strengths and weaknesses on the productivity
dimension, and also retain the element of congruence with the overall
objective of firm profitability. The approach described here is in this
spirit. It has been employed, on an experimental basis, to assess
productivity in varied contexts ranging from mining operations to
power generation, and even for a chain of fast-food restaurants. But
clearly, a considerable amount of additional work needs to be done
to fine-tune the concepts to match the needs of individual
applications.
REFERENCES
Anthony, R. N., and Dearden, J. (1980). Management Control Systems,

Fourth Edition, Irwin, Homewood, Illinois.
Banker, R. D. (1980a). Studies in Cost Allocation and Efficiency Evaluation,
unpublished Ph.D. dissertation, Harvard University, Graduate School of
Business Administration.
Banker, R. D. (1980b). A Game Theoretic Approach to Measuring
Efficiency, Eur. j. Oper. Res., 262-266.
Banker, R. D. (1984). Estimating most productive scale size using data
envelopment analysis, Eur. j. Oper. Res., July, 35-44.
Banker, R. D. (1983). Productivity accounting and variance analysis, working
paper, Carnegie-Mellon University.
Banker, R. D., Charnes, A., and Cooper, W. W. (1984). Estimation of
technical and scale inefficiencies in data envelopment analysis, Manage.
Sci., September, 1078-1092.
Banker, R. D., Charnes, A., Cooper, W. W., and Schinnar, A. (1981) A
biextremal principle of frontier estimation and efficiency evaluation,
Manage. Sci. December, pp. 1370-1382.
Banker, R. D., Charnes, A., Cooper, W. W., and Maindiratta, A. A
comparison of DEA and Translog estimates of production frontiers
using simulated observations from a known technology, forthcoming in

Studies in Productivity Analysis, (Dogramarci, A., ed.), Kliiwer, Boston
(1986).
Banker, R. D., Conrad, R. F., and Strauss, R. P. (1985). A comparative
application of data envelopment analysis and translog methods: An
illustrative study of hospital production, forthcoming in Manage. Sci.
Banker, R. D., and Maindiratta, A. (1983a). Nonparametric Estimation of
Production Frontiers, Working Paper, Carnegie-Mellon University.
Banker, R. D., and Maindiratta, A. (1983b). Loglinear estimation of efficient
production surfaces, submitted to Manage. Sci.
Banker, R. D., and Morey, R. (1983). Efficiency analysis for exogeneously
fixed inputs and outputs, (submitted to Oper. Res.)
Banker, R. D., and Morey, R. (1984). Data envelopment analysis with
categorical variables, (submitted to Manage. Sci.)
Brayton, G. (1983). Measuring productivity to increase profitability, Ind.
Eng. February, 48-56.
Buehler, V. M., and Krishna Shetty, Y. (1981). Productivity Improvement:
Case Studies of Proven Practice, AMACOM: New York.
Caves, D. W., and Christensen, L. R. (1980). Global properties of flexible
functional forms, Am. Econ. Rev. June, 422-432.
Charnes, A., Cooper, W. W., and Rhodes, E. (1978). Measuring the
efficiency of decision making units, Eur.]. Oper. Res. 429-444.
Charnes, A., Cooper, W. W., and Rhodes, E. (1979). Short communication:
measuring the efficiency of decision making units, Eur.]. Oper. Res.,
339.
Charnes, A., Cooper, W. W., and Rhodes, E. (1981). Evaluating program
and managerial efficiency, Manage. Sci.
Coombe, L. W., and Densmore, R. R. (1982). How Detroit-Edison improves
its productivity, Manage. Acc.
Craig, C. E., and Clark Harris, R. (1973). Total productivity measurement at
the firm level, Sloan Manage. Rev. Spring, 13-28.
Cushing, B. E. (1968). Some observations on Demski's Ex Post accounting
system, Acc. Rev. October, 668-671.
Davis, H. S. (1955). Productivity Accounting, Industrial Research Unit, The
Wharton School of the University of Pennsylvania; reprint edition 1978.
Demski, J. (1967). An accounting system structured on a linear programming
model, Acc. Rev. October 701-712
Demski, J. (1968). Some observations in Demski's Ex Post accounting
system: A reply, Acc. Rev. October, 672-674.
Demski, J. (1968). Variance analysis using a constrained linear model, in
256 RAJIV D. BANKER
Studies in Cost Analysis, (Solomons, D. ed.), Irwin, Homewood,

Illinois. pp. 526-540.
Donkersloot, R., Jr. (1981). Productivity through manufacturing control,
Manage. Ace. December.
Fare, R., and Lovell, C. A. K. (1978). Measuring the technical efficiency of
production,]. Econ. Theory, 150-162.
Farrell, M. J. (1957). The measurement of productive efficiency, j. R. Stat.
Soc. Series A, 253-290.
Farrell, M. J., and Fieldhouse, M. (1962). Estimating efficient production
frontiers under increasing returns to scale, ]. R. Stat. Soc., Series A,
252-267.
Gantze, M. E. (1976). in Improving Productivity through Industry and
Company Measurement, Series 2, The National Center for Productivity
and Quality of Life; Washington DC.
Gold, B. (1979). Productivity, Technology and Capital: Economic Analysis,
Management Strategies and Government Policies, Lexington Books,
Lexington Massachusetts.
Gold, (1980). Practical productivity analysis for management accountants,
Manage. Acc. May.
Greenberg, (1973). A Practical Guide to Productivity Measurement, Bureau
of National Affairs: Washington DC.
Greene, W. M. (1980). Maximum likelihood estimation of econometric
frontier functions, j. Econometrics May.
Griliches, Z. (1980). R&D and the productivity slowdown, Am. Econ. Rev.
May, 343-348.
Hayes, R. H. (1982). A Note on Productivity Accounting, Harvard Business
School, report No. 0-682-084.
Hildenbrand, W. (1981). Short-run production functions based on micro-
data, Econometrica September, 1095-1125.
Kaplan, R. (1983). Manufacturing performance: A new challenge for
accounting and management research, Ace. Rev, October, 686-705.
Kendrick, J. W., and Creamer, D. (1965). Measuring Company Productivity:
Handbook with Case Studies, Revised Edition, The Conference Board:
New York.
Kopp, R. J., and Diewert, W. E. (1982). The decomposition of frontier cost
function deviations into measures of technical and allocative efficiency,
]. Econometrics 19,319-331.
Mammone, J. L. (1980a). Productivity measurement: A conceptual overview,
Manage. Ace. June, 36-42.
Mammone, J. L. (1980b). A practical approach to productivity measurement,
Manage. Acc. July, 40-44.
Mensah, Y. (1982). A dynamic approach to the evaluation of input-variable

cost center performance, Ace. Rev. October, 681-700.
Moore, B. (1975). A Plant-Wide Productivity Plan in action: Three Years of
Experience with the Scanlon Plan, National Center for Productivity and
Quality of Working Life: Washington DC.
Moore, B., and Ross, T. L. (1978). The Scanlon Way to Improved
Productivity: A Practical Guide, Wiley, New York.
Richardson, P. R., and Gordon, J. R. M. (1980). Measuring total manufac-
turing performance, Sloan Manage. Rev. Winter, 47-58.
Richmond, J. (1974). Estimating the efficiency of production, Int. Econ.
Rev. 515-521.
Shephard, R. W. (1953). Cost and Production Functions, Princeton Univer-
sity Press, Princeton, New Jersey.
Shephard, R. W. (1970). The Theory of Cost and Production Functions,
Princeton University Press, Princeton, New Jersey.
Varian, H. (1982). The nonparametric approach to demand analysis,
Econometrica, 945-973.
Varian, H. (1984). The nonparametric approach to production analysis,
Econometrica, 579-597.
Wait, D. J. (1980). Productivity measurement: A management accounting
challenge, Manage. Ace. May, 24-30.
CHAPTER 10
THE 1983 GLOBAL

MANUFA CTURING FUTURES
STUDY PRELIMINARY
REPORT SYNOPSIS
Jack Romanosky
1. INTRODUCTION
Senior managers from over 600 business units from North

America, Europe, and Japan participated in the 1983 Global Manu-
facturing Futures Study. The Manufacturing Futures Survey was
initiated, under partial sponsorship by Arthur Andersen & Co.,
through The Manufacturing Roundtable and Boston University.
Professor Jeffrey G. Miller, Boston University, coordinated the
survey, its analysis and conclusions, and its initial presentation at the
Annual Manufacturing Futures Conference held in Boston in April.
This chapter represents a synopsis of the second annual survey's
preliminary findings, which were completed in March 1983.
The purpose of this synopsis is to provide an overview of the
common themes that emerged from the 1983 survey. The survey
provides a broad prospective across the manufacturing environments
in each of three major areas of the world: North America, Japan, and
Europe. This synopsis is intended to provide an overview of the
preliminary observations which have been obtained from the analysis
Jack Romanosky • Arthur Andersen & Co., Five Penn Center Plaza, Philadelphia,
Pennsylvania 19103.
259
260 JACK ROMANOSKY
of survey responses to date. Further analysis of the data is already

underway to explore the "manufacturing futures" of particular
industries within each segment of the world surveyed.
2. CONCLUSIONS OF THE SURVEY
The main finding of the survey is that manufacturers from

North America, Japan, and Europe share a remarkably similar set of
strategic objectives, but a remarkably dissimilar set of manufacturing
plans for achieving these ends. The similarity in ends and the
differences in means between these producers from different parts of
the globe reflect substantially different cost structures, organizational
settings, concerns, and competitive priorities. North Americans seem
to be developing North American solutions, given their current
position, that are quite different from those of their Japanese and
European counterparts. Overall, the Japanese are concentrating on
hardware (new process technologies) to improve their operations,
while, relatively speaking, the North Americans are concentrating on
software and MIS. European firms in general appear to be taking a
position between these two extremes.
It is worth emphasizing that these survey results represent the
summation of the individual plans, concerns, and strategies of a
number of individual business units within each area of the globe.
There are North American firms with profiles remarkably consistent
with the overall Japanese or European profile. There are individual
Japanese firms with plans and concerns that are identical to the
North American or European profiles. The overall survey results
merely suggest the shape of the overall picture emerging from each
part of the world. Doubtless, the difference in these shapes will raise
the question of which is the best one.
3. MAJOR OBSERVATIONS
Major insights from the aggregated survey data come from an

examination of differences across sections of the globe analyzed in
THE 1983 GLOBAL MANUFACTURING FUTURES STUDY 261
terms of strategy and competitive priorities, key concerns, current

practice, and plans for the future.
3.1. Strategies and Competitive Priorities

There was remarkable consistency across the three country
groups in terms of the order of importance of the top four strategic
goals of their business units. In rank order of importance these
shared objectives were as follows:
• Developing new products for existing markets;

• Increasing market share in existing markets;
• Developing new products for new markets;
• Developing new markets for existing products.
This ranking, as well as other data in the survey, suggests there is

a high focus on new product development around the globe.
Differences in emphasis across the globe are apparent in other
strategies selected. For example:
• Both North American and European respondents attached

much higher priority to divestiture and acquisition.
• The Japanese respondents attached a much higher level of
importance to vertical integration, especially backward in-
tegration into sources of supply.
• The very high materials costs of Japanese producers, and the
emphasis the Japanese firms place on price competition, may
explain the attractiveness of this strategic focus to them.
The similarity of the three parts of the world in terms of their

most important strategic directions and objectives suggests that they
view their opportunities in much the same way. However, an analysis
of their strategic priorities suggests that they plan to achieve these
ends in very different ways. Table I contains the rank orderings
of the competitive priorities that each group believes are most
important in competing over the next five years.
Table I shows that, overall, the North American manufacturers
262 JACK ROMANOSKY
TABLE I
Competitive Priorities
North America Japan Europe

Performance/ service: Price/speed-new products Quality / price:
Consistent quality Low price Consistent quality

High-performance Rapid design changes On-time delivery
products
On-time delivery Consistent quality High-performance
products
Fast deliveries High-performance Low price
products
After sale service On-time delivery Rapid design changes
Low price Rapid volume changes Fast deliveries
Rapid design changes After sale service Rapid volume changes
Rapid volume changes Fast deliveries After sale service
consider product quality, performance, and service (e.g., on time

deliveries, fast deliveries; after sale service) to be the most critical to
them. In contrast, the Japanese manufacturers ranked low price, the
ability to make rapid design changes and introduce new products
quickly, followed by consistent quality, as the most important
priorities. The European respondents' priorities fell between these
two extremes.
The North Americans' emphasis on quality in comparison to the
Japanese may reflect the widely acknowledged strength of the
Japanese producers in producing goods of consistent quality in
comparison to the North Americans. The Japanese producers may
not believe they have to focus on their strength as much as the North
Americans do on their weakness.
The Japanese's higher emphasis on low price may well reflect their
end objective, with high quality being the means to that end. The
number one concern of the average Japanese business unit is
producing to high quality standards; the number two concern is
yields. High on their list of recent efforts to improve manufacturing
effectiveness are quality improvement programs.
The North American focus on the service dimensions of
competition and the Japanese focus on price may reflect their
individual emphasis on those nonquality areas where they each have

some relative advantage. But, they must each fight to retain their
advantage:
• The Japanese are fighting other Asian countries with lower

wage costs and other Japanese producers that compete on
pnce .
• The North Americans must protect the advantage offered by
their distribution systems against the encroachment of foreign
producers.
Both the Japanese and North Americans seem to consider the

introduction of new products as a key strategic focus, but only the
Japanese seem to be planning to use the speed of new product
introduction as an important tool for gaining competitive advantage.
3.2. Major Concerns

The chief concerns of the producers from each area of the world
surveyed are shown in Table II.
North American firms appear to be primarily concerned with
areas closely linked to their competitive priorities and! or chief
sources of competitive disadvantage. For example, producing to high
quality standards and introducing new products on schedule are high
on their list. The North Americans seem to be preoccupied with high
overhead costs and low indirect labor (including white collar)
productivity. These concerns are consistent with the finding that
North American overhead costs as a percent of total manufacturing
costs are the highest of the three areas of the world surveyed.
The Japanese producers appear to be highly concerned about
protecting their edge in quality and process technology. Japanese
firms view their quality programs as being consistent with their
ability to produce at low cost and with little waste. Thus, the high
level of concern given to quality and process issues is consistent with
the Japanese competitive focus on low prices, and relative disadvan-
tage in terms of material costs. Also consistent with their emphasis
on speedy design changes and new product introductions is their
264 JACK ROMANOSKY
TABLE II
Major Concerns (Highest First)
Producing to high Producing to high Highlrising overhead

quality standards quality standards costs
High/ rising inventories Yield problemslrejects Producing to high
quality standards
Excess capacity Available qualified Low indirect labor
supervisors productivity
Introducing new products Introducing new products High/rising material
on schedule on schedule costs
Low indirect productivity Falling behind in process Introducing new products
technology on scedule
High/rising inventories High/rising material Poor sales forecasts
costs
High/ rising material Inability to deliver Low direct labor
costs on time productivity
Poor sales forecasts High/rising overhead Excess manufacturing
costs capacity
Low direct labor Low direct labor High/rising inventories
productivity productivity
Yield problems/ rejects Aging work force Impact of government
regulations
Availability of
qualified workers
concern with introducing new products on schedule. The de-

mography and changing social climate in Japan may account for their
concerns with an aging work force, and with the availability of
qualified supervisors and workers.
The profile of concerns of European producers in the survey is
somewhat more difficult to relate to their strategic directions and
competitive priorities. Like their counterparts in North American,
they too seem to be focused on overhead costs and low indirect labor
(and white collar) productivity-in fact, they rank these concerns
even higher than the North Americans. Their overhead costs, while
higher than those of the Japanese, are not quite as high as the North
Americans'. Perhaps this is one area where they are at liberty to do
the most. Their concern with quality and new product introductions
is consistent with their stated strategies and priorities.
A final difference in concerns is seen in the question relating to

objectives in scheduling production. Capacity utilization in North
America only has about one fourth of the relative weight that it has
in Japan and Europe. Similarly, the Japanese emphasis on leveling
production differences is likely to be manifested in the approaches to
production/inventory control systems taken by each.
3.3. Current Practices

One way to understand the differences among the three areas of
the world is to look at what they have been doing recently to
improve manufacturing effectiveness. Table III provides the ten most
popular programs or activities utilized in the prior year to improve
manufacturing effectiveness in each of the three geographical areas.
Table III indicates that many of the programs are common to
each area and most are common to two of the three. Japanese
TABLE III
Current Practices to Impove Effectiveness
Productionlinventory Quality circles Production/inventory

control systems control
Supervisor training Automating jobs Automating jobs
Reducing size of New processes/old Supervisor training
work force products
Integrating manufacturing Reducing size of Manufacturing
MIS work force reorganization
Worker safety Worker safety Making existing systems
work better
Quality circles Reconditioning New processes/old
physical plants products
Direct labor motivation Direct labor motivation Direct labor motivation
Developing new processes Giving workers a broader Changing labor/management
for old products range of tasks relationship
Improved maintenance Value analysis/product Integrating manufacturing
redesign MIS
Making existing systems Manufacturing Worker safety
work better reorganization
266 JACK ROMANOSKY
production/inventory control systems are ranked twelfth (off the

chart). Moreover, other programs that are related, such as integrated
manufacturing MIS, training in production/inventory control
systems, and purchasing also have much lower rankings relative to
Europe and North America. Interestingly, production/inventory
control systems are on the Japanese list of most effective programs. It
would seem that some firms have made this approach work well. The
unique items in the top ten for each area are
• North America:
Improved maintenance;
• Japan:
Giving workers a broader range of tasks;
Reconditioning physical plants;
Value analysis/product redesign;
• Europe:
Changing labor/management relations.
In each case these programs were much lower in ranking for the
other geographical areas.
In the survey data, answers are provided to questions of which
programs were most and least emphasized and most/least effective.
In looking at the lists of most emphasized programs in each area, all
three include manufacturing reorganization and developing new
processes for new products. A major difference is that the European
list includes a large number of programs that are behavioral in nature
and involve changes in work methods. This distinction is even more
dramatic when looking at the list of least effective programs (not
presented in this synopsis). Virtually every item on the European
least effective list involves behavioral and work place changes. The
conclusion is that European respondents have had a much harder job
in this area than have their North American and Japanese
counterparts.
3.4. Programs for the Future

An important part of the survey addressed the plans that the
producers were making to improve their overall operational
TABLE IV
Future Programs to Improve Effectiveness
Production/inventory Automating jobs Automating jobs

control systems
Supervisor training Production/ inventory Direct labor motivation
control systems
Integrating Office automation New processes/new
manufacturing MIS products
Direct labor motivation New processes/new CAM
products
CAM Flexible manufacturing Production/inventory
systems control systems
New processes/old Reducing size of Supervisor training
products work force
New processes/new Introducing robots Integrating
products manufacturing MIS
Quality circles Value analysis/product Quality circles
redesign
Making existing system Giving workers broader Office automation
work better range of tasks
Integrating MIS across Reconditioning physical New processes/old
functions plants products
effectiveness in manufacturing. Table IV contains a list of the ten

most frequently planned programs/activities on which each of the
three country groups is planning to place significant emphasis in the
future. Again, the differences are significant and suggestive of how
these manufacturers are planning to address their concerns and
achieve their strategic goals.
The North American respondent group appears to be placing the
most emphasis on developing and improving its operations in the
following priority:
• Developing and improving MIS systems:

Production and inventory control systems;
Integrating manufacturing MIS;
Integrating manufacturing MIS with other functions;
268 JACK ROMANOSKY
• Development of new process technologies:

Computer assisted manufacturing;
New processes for old and new products;
• Development of people and a better organizational process:
Supervisor training;
Direct labor motivation programs;
Quality circles.
The dominant focus on MIS in North America may be one

reflection of the concern with high overhead costs and white collar
productivity. It may also reflect their service-oriented strategies.
Effectively introducing new products on schedule, rapid deliveries,
and good after sales service require the effective coordination and
communication which good MIS can provide.
The Japanese respondent group is focused in the following
priority areas:
• Development of new product and process technologies:

Automating jobs;
New processes for new products;
Reconditioning physical plants;
Flexible manufacturing systems;
Robots;
Value analysis;
Product redesign;
• Development of production and inventory control systems;
• Development of office automation.
The Japanese focus on automation (and their parallel emphasis

on reducing the size of their work force) is consistent with their
avowed emphasis on price competition. Automation is widely
credited with both decreasing direct labor costs, increasing the
consistency of the quality of products that are produced, and
resulting in higher yields (lower material costs). Their focus on
automation may be the way they see to attack these three problem
areas/objectives at the same time. A further reason for automation is
their demographic concern with the availability of labor (supervisory
and direct) and an aging work force.
Although the emphasis on MIS is not as broad as that of their

North American counterparts, the increased proportion of Japanese
firms that are working on production control systems and office
automation is apparent. Their increasing attention to these MIS
related areas may be a reflection of their interest in developing
communications systems that will allow them to introduce new
products and make design changes more rapidly.
The European respondents' plans for the future seem to again
reflect a middle ground between those of the North American and
Japanese. The Europeans appear to be placing somewhat more
emphasis on automation (automating jobs, new processes for old and
new products, CAM) than the North Americans, but not so much as
the Japanese. The Euorpeans are placing broader emphasis than the
Japanese on MIS, but not so much as the North Americans. Like
both the North Americans and the Japanese, the Europeans have a
sprinkling of programs for developing their people and organiza-
tional processes (direct labor motivation, quality circles, and supervi-
sor training) and appear to be planning to engage in these areas
slightly more than do their counterparts.
One interesting way to view the respondents' programs for the
future is to compare them with past practice. A matching of Table III
with Table IV shows all three areas of the world now interested in
the development of new processes for new products. North America
and Europe both have also added CAM to their list of top ten rated
programs. Office Automation is on both the European and Japanese
lists. North Americans have emphasized office automation in the
past, and now have it on their list of least effective programs. Perhaps
the Japanese and Europeans should take warning. A similar warning
to the Europeans may be in order with regard to quality circles. They
have not had large success with quality circles and other behavioral!
work place change programs in the past.
The Japanese list of future top ten programs has five newcomers
compared to four for Europe and three for North America. As
noted, the strong emphasis is on automation, work force reduction,
and new processes. An interesting question is the extent to which
these firms can change program directions. Are the Japanese firms
more able to rapidly take on this many new programs to improve
manufacturing effectiveness?
270 JACK ROMANOSKY
4. ENVIRONMENTAL OBSERVATIONS
In comparison to other external indicators, the survey results

suggest that the 1982 recession has impacted all three areas of the
world as far as profit margins are concerned. Even so, the same
external indicators suggest that the U.S. and Japanese respondents, as
a group, fared somewhat better than the total population of firms in
their respective countries. Beyond this, however, the overall financial
performance of the participating business units from each area
showed significant differences:
• The North American sample showed the greatest proportion

of firms with high before-tax profit margins (greater than
10%).
• The European sample had the greatest proportion with
negative margins (over 30%).
• The Japanese firms were very highly clustered (80% of the
sample) in the 0%-10% profit margin range.
• The relatively low margins of the Japanese firms m com-
parison with North American firms have been reported
elsewhere. ':.
• The equally good performance on growth in return on assets
that the Japanese firms report suggests that perhaps they make
up in capital utilization and volume what they sacrifice in
margm.
Other survey respondent environmental comparisons include the

following:
• Organization structure:
Japanese respondent firms were typically organized as
operating companies;
North Americans as divisions;
Europeans as holding companies or conglomerates.
". Fortune, May 3, 1983.

• Background of business unit general manager:

Japanese managers have prior experience in manufacturing,
R & D/engineering, and sales in that order. Much more
likely to have had manufacturing experience than North
American or European counterparts.
North American and European managers are more likely to
have had marketing or finance experience than their J ap-
anese counterparts.
• Functional areas currently most likely to get the projects they
propose approved:
Japan:
R & D/product engineering;
Process engineering;
Manufacturing;
North America and Europe;
Manufacturing;
R & D/product engineering;
Marketing.
• Management style:
All three parts of the world judge their top management
groups to be equally cooperative in their work.
The Japanese respondents viewed the decision-making style
of the general manager of their business units to be much
more autocratic than participative, when compared to the
responses of the North American and Europeans.
This unexpected result may reflect differences in norms in
various countries, as Japanese managers are commonly
regarded as being more participative than North American
and European managers.
• Product concentration:
J apanese-consumer goods;
North Americans and Europeans-industrial goods.
• Value added per employee:
Japan and North America-$90,000;
Europe-$51,000.
• Cost structure:
272 JACK ROMANOSKY
Japan:
High concentration of material costs and energy costs
(67% of manufacturing costs);
Direct labor costs (14% of manufacturing costs);
Overhead costs (19% of manufacturing costs);
North America:
Materials and energy costs (60%);
Direct labor costs (13%);
Overhead costs (27%);
Europe:
Materials and energy (59%);
Direct labor (19%);
Overhead (22%).
These cost structure differences may reflect some local reporting

differences. Notwithstanding, they suggest that manufacturers in each
location have substantially different challenges to address. Which cost
areas will provide them with the greatest competitive advantage?
Which are the most out of line with those of other sections of the
globe? Finally, which are most out of line with those of other
sections of the globe? Finally, which are most subject to control?
• Product management
More than 40% of the North American and Japanese
respondents had greater than 20,000 active part numbers.
Only 25% of the European respondents dealt with this
many.
All three areas expect more product customization in the
future. Anticipated change in customization appears to be
highest for the Japanese.
Frequency of engineering changes for North America firm
is significantly higher than for European and Japanese firms.
The average North American respondent processed almost
twice as many engineering changes as the Japanese and
Europeans.
PART IV
TECHNOLOGY MANAGEMENT
AND ORGANIZATION
CHAPTER 11
ORGANIZATIONAL ISSUES
IN THE INTRODUCTION
OF NEW TECHNOLOGIES
Ralph Katz and Thomas J. Allen
1. INTRODUCTION
More than ever before, organizations competing m today's

world of high technology are faced with the challenges of "dualism,"
that is, functioning efficiently today while planning and innovating
effectively for tomorrow. Not only must these organizations be
concerned with the success and market penetration of their current
product mix, but they must also be concerned with their long-run
capability to develop and incorporate in a timely manner the most
appropriate technical advancements into future product offerings.
Research and development-based corporations, no matter how they
are organized, must find ways to internalize both sets of concerns.
Now it would be nice if everyone in an organization agreed on
how to carry out this dualism or even agreed on its relative merits.
This is rarely the case, however, even though such decisions are
critically important to a firm competing in markets strongly affected
by changing technology (Allen, 1977; Roberts, 1974). Amidst the
pressures of everyday requirements, decision makers representing
different parts of the organization usually disagree on the relative
Ralph Katz. Graduate School of Business, Northeastern University, Boston, Mas-

sachusetts 02115. Thomas J. Allen. Sloan School of Management, Massachusetts
Institute of Technology, Cambridge, Massachusetts 02139.
275
276 RALPH KATZ AND THOMAS]. ALLEN
wisdom of allocating resources or particular RD&E talents among

the span of technical activities that might be of benefit to today's
versus tomorrow's organization. Moreover, there are essentially no
well-defined principles within management theory on how to struc-
ture organizations to accommodate these two sets of conflicting
challenges. Classical management theory with its focus on scientific
principles deals only with the efficient production and utilization of
today's goods and services. The principles of high task specialization,
unity of command and direction, high division of labor, and the
equality of authority and responsibility all deal with the problems of
structuring work and information flows in routine, predictable ways
to facilitate production and control through formal lines of authority
and job standardization. What is missing is some comparable theory
that would also explain how to organize innovative activities within
this operating environment such that creative, developmental efforts
will not only take place but will also become more accepted and
unbiasedly reviewed, especially as these new and different ideas begin
to "disrupt" the smooth functioning organization. More specifically,
how can one structure an organization to promote the introduction
of new technologies and, in general, enhance its longer-term innova-
tion process, yet at the same time, satisfy the plethora of technical
demands and accomplishments needed to support and improve the
efficiency and competitiveness of today's producing organization?
Implicit in this discussion, then, is the need for managers to
learn how to build parallel structures and activities that would not
only permit these two opposing forces to coexist but would also
balance them in some integrative, meaningful way. Within the RD&E
environment, the operating organization can best be described as an
"output-oriented" or "downstream" set of forces directed towards
the technical support of the organization's current products and
towards getting new products out of development and into manufac-
turing or into the marketplace. Typically, such pressures are control-
led through formal structures and through formal job assignments to
project managers who are then held accountable for the successful
completion of product outputs within established schedules and
budget constraints.
At the same time, there must be an "upstream" set of forces that
ORGANIZATIONAL ISSUES 277
are less concerned with the specific architectures and functionalities

of today's products but are more concerned with the various core
technologies that might underlie the industry or business environ-
ment not only today but also tommorow. They are, essentially,
responsible for the technical health and excellence of the corporation,
keeping the company up-to-date and technically competitive in their
future business areas.
In every technology-based organization, as discussed by Katz
and Allen (1985), the forces that represent this dualism compete with
one another for recognition and resources. The conflicts produced by
this competition are not necessarily harmful; in fact, they can be very
beneficial to the organization in sorting out project priorities and the
particular technologies that need to be monitored and pursued,
provided there are mechanisms in place to both support and balance
these two forces.
If the product-output or downstream set of forces becomes
dominant, then there is the likelihood that sacrifices in using the
latest technical advancements may be made in order to meet budget,
schedule, and immediate market demands. Given these pressures,
there are strong tendencies to strip the organization of its research
activities and to deemphasize longer-term, forward-looking techno-
logical efforts and investigations in order to meet current short-term
goals which could, thereby, mortgage future technical capabilities.
Under these conditions, requirements for the next generation of new
product developments begin to exceed the organization's in-house
expertise, and product potentials are then oversold beyond the
organization's technical capability.
At the other extreme, if the research or upstream technology
component of the organization is allowed to dominate development
work within R&D, then the danger is that products may include not
only more sophisticated but also perhaps less proven, more risky, or
even less marketable technologies. This desire to be technologically
aggressive-to develop and use the most attractive, most advanced
technology-must be countered by forces that are more sensitive to
the operational environments and more concerned with moving
research efforts into some final physical reality. Technology is not an
autonomous system that determines its own priorities and sets its
278 RALPH KATZ AND THOMAS J. ALLEN
own standards of performance. To the contrary, market, social, and

economic considerations eventually determine priorities as well as the
dimensions and levels of performance necessary for successful
commercial application (Utterback, 1974).
To balance this dualism-to be able to introduce the new
technologies needed for tomorrow's products while functioning
efficiently under today's current technological base, is a very difficult
task. Generally speaking, the more the organization tries to operate
only through formal mechanisms of organizational procedures,
structures, and controls, the more the organization will move
towards a functioning organization that drives out its ability to
experiment and work with new technological concepts and ideas.
More informal organizational designs and processes are therefore
needed to influence and support true innovative activity, countering
the organization's natural movement towards more efficient produc-
tion and bureaucratic control. These informal mechanisms are also
needed to compensate for the many limitations inherent within
formal organizational structures and formal task definitions. In the
rest of this chapter, we will describe three general areas of informal
activity that need to take place within an RD&E environment (in
parallel with the formal, functioning organization) in order to
enhance the innovation process for the more timely introduction of
new technologies into the corporation's product portfolio. The
general proposition is that these areas of informal activity need to be
managed within the RD&E setting, strengthening and protecting
them from the pressures of the "productive" organization in order to
increase the organization's willingness and ability to deal with the
many advancements that come along, especially with respect to new
areas of technology.
2. PROBLEM SOLVING, COMMUNICATIONS, AND THE

MOBILITY OF PEOPLE
To keep informed about relevant developments outside the

organization as well as new requirements within the organization,
R&D professionals must collect and process information from a large
variety of outside sources. Project members rarely have all the

requisite knowledge and expertise to complete successfully all of the
tasks involved in new technical innov:ltions; information and assis-
tance must be drawn from many sources beyond the project both
within and outside the organization. Furthermore, if one assumes
that the world of technology outside the organization is larger than
the world of technology inside the organization, then one should also
expect a great deal of emphasis within R&D on keeping in touch
with the many advancements in this larger external world. Allen's
(1977) 20 years of :·esearch work on technical communications and
information flows clearly demonstrates just how important this
outside contact can be in generating many of the critical ideas and
inputs for more successful research and development activity.
At the same time, the research findings of many studies,
including Katz and Tushman (1981), Allen (1977), and Pelz and
Andrews (1966), have consistently shown that the bulk of these
critical outside contacts comes from face-to-face interactions among
individuals. Interpersonal communications rather than formal techni-
cal reports, publications, or other written documentation are the
primary means by which engineering professionals collect and
transfer important new ideas and information into their organizations
and project groups. In his study of engineering project teams, for
example, Allen (1977) carefully demonstrated that only 11 % of the
sources of new ideas and information could be attributed to written
media; the rest occurred through interpersonal communications.
Many of these "creative" exchanges, moreover, were of a more
spontaneous nature in that they arose not so much out of formal
project requirements and interdependencies but out of factors
relating to past project experiences and working relationships, the
geographical layouts of office locations and laboratory facilities,
attendances at special organizational events and social functions,
chance conversations with external professionals and vendors at
conferences and trade shows, and so on. Anything that can be done
to stimulate informal contacts among the many parts of the organiza-
tion and between the organization's R&D professionals and their
outside technology and customer environments is likely to be helpful
in terms of both technology development and technology transfer.
Since communication processes· play such an important role in

fostering the creative work activities of R&D members, it would be
nice if each individual or project team were naturally willing or
always motivated to expose themselves to fresh ideas and new points
of view. Unfortunately, this is usually not the case as engineering
individuals continue to work in a particular project area or in a given
area of technology. In fact, one of the more important assumptions
underlying human behavior within organizations is that people are
strongly motivated to reduce uncertainty (Katz, 1982). As part of this
process, individuals, groups, and even organizations strive to struc-
ture their work environments to reduce the amount of stress they
must face by directing their activities and interactions toward a more
predictable level of certainty and clarity. Over time, then, engineers
and scientists are not only functioning to reduce technical un-
certainty, they are also functioning to reduce their "personal and
situational" uncertainty within the organization (Katz, 1980). In the
process of gaining increasing control over their task activities and
work demands, three broad areas of biases and behavioral responses
begin to emerge. And the more these trends are allowed to take place
and become reinforced, the more difficult it will be for the
organization to consider seriously the potential, long-term advan-
tages of the many new and different technologies that are slowly
being developed and worked on by the larger outside R&D
community.
2.1. Problem-Solving Processes
As R&D professionals work together in a given area for a long

period of time and become increasingly familiar with their work
surroundings, they become less receptive toward any change or
innovation that threatens to disrupt significantly their comfortable
and predictable work patterns of behavior. In the process of reducing
more and more uncertainty, these individuals are likely to develop
routine responses for dealing with their frequently encountered tasks
in order to ensure predictabiiity, coordination, and economical
information processing. As a result, there develops over time
increasing rigidity in their problem-solving activities-a kind of
functional stability that reduces their capacity for flexibility and

openness to change. Behavioral responses and technical decisions are
made in fixed, normal patterns; and consequently, new or changing
situations that may require technical strategies that do not fit prior
problem-solving molds are either ignored or forced into these
established molds. R&D professionals interacting over a long period,
therefore, develop work patterns that are secure and comfortable,
patterns in which routine and precedent play a relatively large part.
They come, essentially, to rely more and more on their customary
ways of doing things to complete project requirements. In their
studies of problem-solving strategies, for example, Allen and Marquis
(1963) show that within R&D there can be a very strong bias for
choosing those technical strategies and approaches that have worked
in the past and with which people have gained common experience,
familiarity, and confidence; all of which inhibit the entry of
competing tactics involving new technologies, new ideas, or new
competencies.
What also seems to be true is that as engineers continue to work
in their well-established areas of technology and develop particular
problem-solving procedures, they become increasingly committed to
these existing methods. Commitment is a function of time, and the
longer individuals are asked to work on and extend the capabilities of
certain technical approaches, the greater their commitment becomes
toward these approaches. Furthermore, in accumulating experience
and knowledge in these technical areas, R&D has often had to make
clear presentations, showing progress and justifying the allocation of
important organizational resources. As part of these review proces-
ses, alternative or competing ideas and approaches were probably
considered and discarded, and with such public refutation, com-
mitments to the selected courses of action become even stronger.
Individuals become known for working and building capability in
certain technical areas, both their personal and organizational identi-
ties become deeply ensconced in these efforts, and as a result, they
may become overly preoccupied with the survival of their particular
technical approaches, protecting them against new technical alterna-
tives or negative evaluations. All of the studies that have retrospec-
tively examined the impact of major new technologies on existing
organizational decisions and commitments arrive at the same general

conclusion: those working on and committed to the old, invaded
technology fail to support the radical new technology; instead, they
fight back vigorously to defend and improve the old technology (e.g.,
Cooper and Schendel, 1976; Schon, 1963). And yet, it is often these
same experienced technologists who are primarily asked to evaluate
the potential effects of these emerging new technologies on the future
of the organization's businesses. It is no wonder, therefore, that in
the majority of cases studied, the first commercial introduction of a
radical new technology has come from outside the industry's
traditional competitors.
2.2. Communication and Information Processing

One of the consequences of increased behavioral and technical
stability is that R&D groups also become increasingly isolated from
outside sources of relevant information and important new ideas. As
engineers become more attached to their current work habits and
areas of technical expertise, the extent to which they are willing or
even feel they need to expose themselves to new ideas, approaches, or
technologies becomes progressively less and less. Instead of being
vigilant in seeking information from the outside world of technology
or from the market place, they become increasingly complacent
about external events and new technological developments. After
studying the actual communication behaviors of some 350 engineer-
ing professionals in a major R&D facility, Katz and Allen (1982)
found that as members of project teams worked together, gained
experience with one another, and developed more stable role
assignments and areas of individual contribution, the groups also
communicated less frequently with key sources of outside informa-
tion. Research groups, for example, failed to pay sufficient attention
to events and information in their external R&D community while
product development and technical support groups had reduced
levels of communication with their internal engineering colleagues
and with their downstream client groups from marketing and
manufacturing. Such low levels of outside interaction also result in
stronger group boundaries, creating tougher barriers to effective
communication and more difficult information flows not only among

R&D groups but also to other organizational divisions and to other
areas outside the organization.
Another set of forces that affects the amount and variety of
outside contact that R&D employees may have is the tendency for
individuals to want to communicate only with those who are most
like themselves, who are most likely to agree with them, or whose
ideas and viewpoints are most likely to be in accord with their own
interests and established perspectives. Over time, R&D project
members learn to interact selectively to avoid messages and informa-
tion that might conflict with their current dispositions toward
particular technologies or technical approaches, thereby, restricting
their overall exposure to outside views and allowing themselves to
bias the interpretation of their limited outside data to terms more
favorable to their existing attitudes and beliefs. Thus, the organiza-
tion ends up getting its critical and evaluative information and
feedback not from those most likely to challenge or stretch their
thinking but from those with whom they have developed comfor-
table and secure relationships, i.e., friends, peers, long-term suppliers
and customers, etc. And it is precisely these latter kinds of
relationships that are least likely to provide the inputs and thinking
necessary to stimulate the organization's movement into new techni-
cal areas.
2.3. Cognitive Processes

One of the dilemmas of building in-house capability in particu-
lar areas of technology is that engineers responsible for the success of
these technical areas become less willing to accept or seek the advice
and ideas of other outside experts. Over time, these engineers may
even begin to believe that they possess a monopoly on knowledge in
their specialized areas of technology, seriously discounting the
possibility that outsiders might be producing important new ideas or
advances that might be of use to them. And if this kind of outlook
becomes mutually reinforced within a given R&D area or project
group, then these individuals often end up relying primarily on their
own technical experiences and know-how, and consequently, are
284 RALPH KATZ AND THOMAS j. ALLEN
more apt to dismiss the critical importance of outside contacts and

pay less attention to the many technical advances and achievements in
the larger external world. It is precisely this attitude, coupled with
the communication and problem-solving trends previously described,
that helps explain why most of the successful firms in a very new area
of technology had never participated in the old or substituted area of
technology.
This rather myopic outlook within R&D is also encouraged as
technologists become increasingly specialized, that is, moving from
broadly defined capabilities and solution approaches to more nar-
rowly defined interests and specialities. Pelz and Andrews (1966)
argue from their study of scientists and engineers that with increasing
group stability, project member preferences for probing deeper and
deeper into a particular technological area becomes greater and
greater while their preferences for maintaining technical breadth and
flexibility gradually decrease. Without new challenges and oppor-
tunities, the diversity of skills and of ideas generated are likely to
become progressively more narrow. They are, essentially, learning
more and more about less and less. And as engineers welcome
information from fewer sources and are exposed to fewer alternative
points of view, the more constricted their cognitive abilities become,
resulting in a more restricted perspective of their situation and a more
limited set of technological responses from which to cope. One of the
many signs of obsolescence occurs when engineers retreat to their
areas of specialization as they feel insecure addressing technologies
and problems outside their direct fields of expertise and experience.
They simply feel more comfortable and creative when they can see
their organizational contributions in terms of their past performance
standards rather than on the basis of future needs and requirements.
Finally, there is not only a strong tendency for technologists to
communicate with those who are most like themselves, but it is just
as likely that continued interaction among members of an R&D
project team will lead to greater homogeneity in knowledge and
problem-solving behaviors and perceptions. The well-known proverb
"birds of a feather flock together" makes a great deal of sense, but it
is just as accurate to say that "the longer birds flock together, the
more of a feather they become." One can argue, therefore, that as
R&D project members work together over a long period, they will
reinforce their common views and commitments to their current
technologies and problem-solving approaches. The group not only
tries to hire or recruit new members like themselves, but they also
begin to attract people like themselves, thereby exacerbating the
trend towards greater homogeneity and consensus and less diversity.
Such shared values and perceptions, created through group interac-
tions, act as powerful constraints on individual attitudes and behav-
iors and provide group members with a strong sense of identity and a
great deal of assurance and confidence in their traditional activities.
At the same time, however, these shared systems of meaning and
beliefs restrict individual creativity into new areas and isolate the
group even further from important outside contacts and technical
developments, thereby, causing the old technologies to become even
more deeply entrenched.
2.4. Mobility of People and the "Not Invented Here" Syndrome

What is implied by all of this discussion is that R&D managers
need to learn to observe the strong biases that can naturally develop
in the way engineers select and interpret information, in their
willingness to innovate or implement radically new technological
approaches, or in their cognitive abilities to generate or work with
new technical options so that appropriate actions can be undertaken
to encourage R&D to become more receptive and responsive to new
ideas and emerging technological opportunities. The trends described
here are observable; one can determine the extent to which project
groups are communicating and interacting effectively with outside
information sources, whether project groups are exposing themselves
to new ideas and more critical kinds of reviews, or whether a project
group is becoming too narrow and homogeneous through its hiring
practices.
In the best-selling book, In Search of Excellence, organizations
are encouraged by Peters and Waterman to practice the Hewlett
Packard philosophy of MBW A (Management by Wandering
Around). But managers have to know what to look for as they
wander around. In particular, technical managers can try to detect the
degree to which these different trends are materializing, for the way
engineering groups come to view their work environments will be
very critical to the organization's ability to introduce and work with
new technologies. The more the perceptual outlook of an R&D area
can be characterized by the problem-solving, informational, and
cognitive trends previously described, the more likely it has in-
ternalized what has become known in the R&D community as the
"Not Invented Here" (NIH) or the "Nothing New Here" (NNH)
syndrome. According to this syndrome, project members are more
likely to see only the virtue and superiority of their own ideas and
technical activities while dismissing the potential contributions and
benefits of new technologies and competitive ideas and ac-
complishments as inferior and weak.
It is also argued here that the most effective way to prevent
R&D groups from developing behaviors and attitudes that coincide
with this NIH syndrome is through the judicious movement of
engineering personnel among project groups and organizational
areas, keeping teams energized and destabilized. Based on the findings
of Katz and Allen (1982), Smith (1970), and several other studies,
new group members not only have a relative advantage in generating
fresh ideas and approaches, but through their active participation,
project veterans might consider more carefully ideas and technologi-
cal alternatives they might otherwise have ignored. In short, project
newcomers represent a novelty-enhancing condition, challenging and
improving the scope of existing methods and accumulated
knowledge.
The mobility of people within the organization is a most fruitful
approach for keeping ideas fresh, building insights, and maintaining
innovative flexibility. Japanese organizations, for example, assume
that the best course of development for capable individuals is lateral
rotation across major functional areas of the firm before upward
advancement takes place. In a Japanese company, an engineer
progressing well may move from R&D into marketing, then into
manufacturing, and perhaps back into R&D at a higher level. This is
seldom the kind of career track that American firms find appropriate;
yet, we all know for sure the kinds of problems one is avoiding as
well as the benefits that would accrue over the long run through the
greater use of rotation programs even if rotation were limited to

between research and development and engineering groups.
In an additional attempt to foster new thinking and to build
stronger intraorganizational bridges and communication networks,
some companies hold special meetings in which organizational areas
report on what they have been doing and on the kind of capability
they have. The 3M Corporation, for example, holds a proprietary
company fair at which there are presentations of technical papers,
exhibits, and demonstrations of projects and prototypes. The fair
enables the rest of the people in the company to begin to learn about
what is taking place in other divisions for laboratories. The Monsanto
Company uses what it calls the Monsanto technical community to
bring together technical people, trained in similar disciplines but
employed in different divisions of the firm, and it convenes these
people in different workshops and groups, encouraging them to
exchange ideas and information. These kinds of programs can be very
helpful in fostering communication and in stimulating the identifica-
tion of new technical capabilities as well as the identification of new
market and technical needs throughout the firm.
3. ORGANIZATIONAL STRUCTURES
Unlike productivity, which is the efficient application of current

solutions, innovation usually connotes the first utilization of a new
or improved product, process, or practice. Innovation, as a result,
requires both the generation or recognition of a new idea followed by
the implementation or exploitation of that idea into a new or better
solution. So far, we have discussed organizational processes to the
extent that they primarily affect the idea-generation phase of the
innovation process. It is just as important, of course, for an
organization to plan for the idea-exploitation phase, where exploita-
tion includes the appraisal, focusing, and transferring of research
ideas and results for their eventual utilization and application. To say
that one is managing or organizing for the introduction of new
technologies within the innovation process implies that one is
"pushing" the development and movement of new technical ideas
288 RALPH KATZ AND THOMAS ]. ALLEN
and capabilities downstream through the organization from research

to development to engineering and even into manufacturing and
perhaps some phase of customer distribution.
Innovation, then, is a dynamic process involving the movement
and transfer of technologies across internal organizational bound-
aries. Formal organizational design, on the other hand, is a static
concept, describing how to organize collections of activities within
well-defined units and reporting relationships, e.g., research, ad-
vanced development, product development, engineering, quality
assurance, etc. Formal organizational structures tell us what to
manage and with whom to interact within certain areas of in-
terdependent activity; they tell us little about how to move informa-
tion, ideas, and in particular technologies across different organiza-
tional areas, divisions, or formal lines of authority. In fact, formal
structures tend to separate and differentiate the various organizational
groupings, making the movement of ideas and technologies par-
ticularly difficult across these groupings, especially if there are no
compensating integrating mechanisms in place. And it is in the
movement of new technological concepts from research to advanced
development to successful product development that we are par-
ticularly interested.
The effective organization, therefore, needs to cause the results
of R&D to be appropriately transferred. Technically successful
R&D, especially if it embraces new radical technologies, is very likely
to pose major problems of linkages with the rest of the firm,
particularly product development, engineering, manufacturing,
marketing, sales, field service, and so on. A company can do a terrific
job of R&D and a terrible job of managing the innovating process
overall simply because the results of R&D have never been fully
exploited and successfully moved downstream. Witness, for example,
the problems of Xerox, where the R&D labs have generated and
surfaced many major new advances and approaches only to discover
that the company has failed to fully exploit and capture benefit from
many of them. Other corporations, on the other hand, have benefited
extremely well from Xerox's research activities-so m.any in fact that
some have quipped that Xerox's research facilities should be declared
a national resource instead of a resource for Xerox (see Fortune
magazine, September, 1983).
Over the past decade or so, Roberts (1979) has been studying the
problems of moving R&D results through the organization. From
carrying out these studies, he has found that most large organizations
have been dissatisfied with the degree of transfer of their own R&D
results and feel very uncomfortable about how little of their good
technical outcomes ever reach the marketplace and generate profit-
able pay-back for the firm. The R&D labs he studied seemed to have
broad enough charters to do almost anything they chose, but ended
up being quite narrow as to what they in fact implemented within
their own organizations. To enhance the transfer of R&D results
across the barriers of organizational structures, Roberts (1979)
advocates the building of bridges; and in particular, he recommends
three different groups of bridges: procedural, human, and
organizational.
The procedural approaches, according to Roberts, try to tie
together both the R&D unit and the appropriate receiving units by
joint efforts. In the case of new technological concepts, the most
immediate receiving unit is typically some advanced development
group or some divisional product development organization that
receives the output from a centralized research and development lab.
The kinds of procedural bridges that have been suggested include
joint planning of R&D programs and joint staffing of projects,
especially immediately before and after transfer, for those are the
most critical phases of the process in which key know-how and
information can easily slip through the cracks.
Joint appraisal of results by research, development, and any
other appropriate downstream unit or customer is also employed in
some labs. From the viewpoint of generating useful information, the
best time to carry out joint appraisal of results is when failure has
occurred, for there is usually something objective to look at from
which one might be able to learn and improve. At the same time,
however, this exercise must be done carefully and sensitively to
prevent this opportunity from becoming a situation of mutual
fingerpointing, showing why the other group is really at fault and
how those people caused the failure. In these joint appraisals, the
attributions of failure should be centered around substantive issues
that can be dealt with behaviorally, structurally, or procedurally;
otherwise, intergroup conflicts and differences will be strengthened,
290 RALPH KATZ AND THOMAS j. ALLEN
which is likely to cause even greater difficulty in future technological

handoffs. Joint appraisal of successes should also not be overlooked,
for they can be very helpful in generating the goodwill and trust
necessary to strengthen organizational linkages, especially after a
history of prior difficulty or failure.
The establishment of human bridges also helps to cope with
transfer issues. Interpersonal alliances and informal contacts in-
evitably turn out to be the basis of integration and intraorganiza-
tional cooperation that really matter. The human approaches focus
on the relationships that convey information between people, that
convey the shift of responsibility from one person to another, and
that convey enthusiasm for the project. Roberts argues strongly, in
fact, that the building of human bridges is by far the best way to
transfer this vital enthusiasm and commitment.
Technology moves through people, and the most effective of
these human bridges is the actual movement of people in two
directions. Upstream movement of development engineers to join the
R&D effort well in advance of the intended transfer is a very
important step. This transfers information from the product develop-
ment areas into the Research process, creates an advocate to bring the
research results downstream, and builds interpersonal ties for the
later assistance that will inevitably be needed as the technology
encounters problems. Downstream movement of research individuals
will also be helpful in providing the technical expertise necessary for
development to build up its own understanding and capability.
In addition to the specific movement of people, human bridges
are also built through the interpersonal communication systems that
have developed over time through the history of working relation-
ships, rotation programs, task force participation, and other or-
ganizational events and activities. Another important device to be
considered is the joint problem-solving meeting in which develop-
ment individuals are asked to sit down with research colleagues to let
them explain their difficulties and initial problem-solving thinking.
Such meetings are not only helpful in dealing with specific project
problems but will also be useful in building stronger human bridges
between the related R&D areas and may even be helpful in solving
additional related problems that were not initially put forth.
The final area for considering the movement of R&D results

towards development and eventual commercialization consists of
organizational changes and organizational bridges. According to
Roberts, these are the toughest kinds to create and implement
effectively in an organization. It is far easier to alter procedures or to
try to build human bridges across groups than it is to change
organizational arrangements and relationships. Nevertheless, several
different structural approaches can be effective under different
organizational conditions. Some organizations have developed spe-
cialized transfer groups, created solely for the purpose of transferring
important technical advances or important new processes. Under this
approach, the transfer group is like the licensor of a technology who
is not just sending equipment and documentation but who is also
responsible for training others to work with the technology, for
installing the equipment, etc. If used, the specialized transfer group
should consist of at least a few of the key technical players. Senior
management should not be allowed to argue that they cannot spare
the superstars of the research organization to support development or
manufacturing engineering.
Another organizational approach is to employ integrators or
integrating groups that are given responsibility for straddling the
various parts of the RD&E organization. This is a very uncomfort-
able and a very difficult job to assume because it is extremely difficult
to ask someone to take care of an integrating function across two
separate sub organizations when he or she does not have respon-
sibility for either the sending or the receiving organization. To
perform this function successfully requires someone who can cope
with the political sensitivities of multiple groups and who has built
substantial informal influence and credibility within the organization.
Finally, a variety of corporate venture strategies can be con-
sidered by companies that are concerned with developing new
technical approaches, new product lines, or want a stronger emphasis
on technical entrepreneurship. Roberts (1980) suggests a large variety
of possible venture strategies, ranging from the high corporate
involvement of internal venturing to low corporate involvement
through venture capital investments in outside firms for the purpose
of gaining windows on technology and new market opportunities.
Additional venture strategies are also described by Roberts, including

the coupling of R&D efforts from both the large corporation and the
small independent firm. In general, there is no single best way to
organize for the effective introduction of new technologies; but the
more informal mechanisms one puts in place to foster both the idea
generation and the idea exploitation phases, the more one is likely to
be successful at managing the innovation process.
4. ORGANIZATIONAL CONTROLS
All of these organizational attempts at stimulating new techno-

logical innovation will fall flat, of course, if organizational controls
are not consistent with the innovation process. In looking at many
case histories of successful versus unsuccessful innovations based on
radical new technologies, Cohen et ai. (1979) and several other
studies have identified a number of factors as being critically
important for trying to influence the generation and successful
movement of new technologies through the organization.
4.1. Technical Understanding

One of the most important issues in working with new
technologies is that the research function must fully understand the
main technical issues of the technology before passing it on. Although
this point seems obvious, it is often overlooked. The research
function must focus not only on the benefit of the new technology in
and of itself; it must also deal with the technology's limitations
relative to conventional technologies and to other new technological
approaches. In the early days of transistors, for example, one large
electronics company spent a great deal of money and many years of
research effort on understanding the materials and processing prob-
lems of germanium for point contacts and junction transistors.
Unfortunately, the research organization failed to compare the use of
germanium to silicon, whose own development was continuing to
make a great deal of progress. Only after many years did the
organization finally realize the limitations in the advantages of
germanium over silicon and these limitations had less to do with the
devices themselves and more to do with device implementation in
packaging and circuitry.
It is also important, therefore, to make sure that research
understands where the new technology might fit in with respect to
the product line or at least what requirements must be met to reach
this fit. Research should not waste its time solving problems that do
not exist or producing technologies that cannot be sold. Whirlpool,
for example, invested substantial research resources in making
applicance motors more energy efficient long before the oil crisis, but
of course, the marketplace was not yet interested in these kinds of
advances. Similarly, GE conducted a great deal of research in
environmental concerns in the 1940s but at that time there was very
little interest in improving the ecology of our environment. As a last
example, DuPont developed Corfam as a synthetic substitute for
leather, but unfortunately for DuPont, the public was perfectly
satisfied with leather and saw no need for the man-made substitute.
Full understanding also means that research must begin to
examine the means of manufacturing, the availability of key materials
and technical talents, the ease of use, and so on. Air Products and
Chemicals, for example, spent millions of dollars to develop a
fluorination process so that textile manufacturers could make fabrics,
especially polyesters, more resistant to oil and grease. Unfortunately,
textile manufacturers did not want fluorine-a poisonous and cor-
rosive gas-anywhere near their plants and refused to buy the
system. Research should also be able to make, at the very least,
preliminary cost estimates. One of the most basic elements of a
technology is its cost. In fact, a study of technology programs at GE
concluded that most of the barriers to the introduction of new
technologies (even hardware and software) were cost constraints and
not technical feasibility; it was getting the technology to perform
capably at a marketable cost.
To help ensure these kinds of requiements, some labs have
begun to hire full-time marketing representatives and cost estimators
as a regular part of the R&D organization. Previously, corporate
R&D organizations were completely dependent on product line
divisions for both marketing and sales effort and for business and
economic analysis as well. These dependencies, especially the latter,

were harmful in getting research projects justified, supported, and
accepted by the divisions who were supposed to be the eventual
customer of the research results.
4.2. Technical Feasibility

All too often, a technology is transferred before there has been
sufficient time within research to demonstrate true feasibility. Such
pressures can come from the downstream organization or they can
arise from the "unbridled enthusiasm" of the researchers themselves.
In either case, it would be more beneficial to discuss what constitute
feasibility and for research to strive to achieve it.
Most new technical concepts do not succeed simply because they
must run a guantlet of barriers as they enter the main part of the
functioning organization. In many cases, the new technology is
embedded within a system of established technologies. The question
then is whether the new technology will offer a sufficient competitive
advantage to warrant its incorporation into this interdependent
system, perhaps changing drastically the tooling and the overall
manufacturing process. Experienced technologists will typically warn
you that what you do not yet know about the workings of a new
technical advance will probably come back to haunt you. What often
appears to be a simple technical issue turns ot to be more complicated
than we realize. GE discovered a fiber, for example, that looked and
behaved more like wool than any synthetic yet known. Unfortu-
nately, the fiber disintegrates in today's cleaning solvents, and the
problem has yet to be solved.
4.3. Research and Development Overlap

As previoulsy discussed, it is very helpful to the movement of a
new technology if development, or some other appropriate receiving
organization, also has a group of technical people who have been
getting up to speed on the technology before the actual transfer, e.g.,
the presence of "ad tech" groups. Such advanced technical activities
within development can greatly aid the movement of technology and
the smoothing of conflicts.
In a similar fashion, it is also important for research to maintain

some activity to support and defend the new technology or to find
new ways to extend the technology. Research must not be allowed to
feel that it is "finished" at the time of transfer, for if this feeling is
present, their willingness and enthusiasm to support the technology
will be minimal. Most new technologies are relatively crude at first.
Ball-point pens, for example, blotted, skipped, stopped writing all
together, and even leaked in consumers' pockets when they first
appeared on the market. The first transistors were expensive and had
sharply limited frequencies, power capabilities, and temperature
tolerances. Such experiences are very typical of new technologies,
especially radical new technologies. And the more prepared research
is to help "push" the technology, the less likely it will be for the new
technology to be dismissed prematurely as a "fad" or as a technology
with very limited application.
4.4. Growth Potential

As a related point, all too often a research program sells itself
short by being too narrow and not showing a clear path towards
technical growth and growth in product applicability. In almost
every instance, when the new technology appears on the scene, the
old technology is forced to "stretch" itself, often with major
advances being achieved in the threatened technology. Under these
circumstances, the new technology is in the position of trying to
chase or catch a "changing target." Moreover, this new potential in
the old technology often holds back the entry of the new technology.
Advances in flashbulbs, for example, held off the widespread use of
electronic flash for quite some time, while advances in magnetic tape
audio and video recording have prevented the emergence of ther-
moplastic recording. In their well-known study of strategic responses
to technological threats, Cooper and Schendel (1976) indicate that in
the majority of cases, sales of the old technology did not decline after
the introduction of a new technology. To the contrary, sales of the
old technology expanded even further. It is for these reasons that the
diffusion and substitution of a radical new technology must be
viewed as a long-term process and research and development must
296 RALPH KATZ AND THOMAS ]. ALLEN
carefully prepare to argue and demonstrate why the pressured

organization should be patient during this time period.
4.5. Organizational Slack and Sponsorship

When an organization pushes too hard for productivity within
the RD&E environment, trying to measure and control all aspects of
the innovation process, there is little room or slack for experimenting
or pursuing novel ideas and concepts. The environment is simply too
tightly run and the climate becomes unfavorable for very new or
long-term innovation. Engineers and scientists become anxious,
restrict the depth of exploration along new paths, and center their
attention upon issues closely related to the company's immediate
output. Creative innovation, on the other hand, is harder to measure
and takes a longer period to assess. It requires speculative in-
vestments on the part of the firm that wants to nurture the ideas and
the experimenting activities that will eventually be worth it.
Given all of the resistance and testing that a new technological
idea will eventually encounter from the functioning organization and
from operational review committees, strong corporate sponsorship is
needed to protect new technological innovations. And the more
radical the new technology, the stronger the corporate sponsorship
has to be. One of the observations we have made from working and
consulting with many technology companies is that most (and in
some high technology companies, all) radical new technologies have
had to have well-identified sponsorship at the corporate level in order
to succeed.
Another important finding from retrospective studies of radical
innovation is that new technologies are not really new! By this, we
mean that technological change is a relatively continuous and
incremental process which casts shadows far ahead. According to
Utterback and Brown (1972), the information incorporated in
successful new innovations has been around for roughly 5-30 years
prior to its use. They further argue that there are many multiple
signals within the external environment that can be used to predict
the direction and impact of future technological changes and
development. Von Hippel (1983), for example, argues that one
can often anticipate future innovations by identifying what he calls

"lead users," that is, users whose needs today foreshadow the needs
of the general marketplace tomorrow. Nevertheless, even if particular
areas of new technology were identified as extremely important,
wihout strong sponsorship it is unlikely that sufficient resources
would be diverted to it, that engineers would be isolated from other
pressures or tasks to work on it, or that they would be given
sufficfficient resources would be diverted to it, that engineers would
be isolated from other pressures or tasks to work on it, or that they
would be given sufficient uninterrupted time to complete it. One of
the reasons why so many new technologies are introduced through
the emergence or spin-offs of new firms is that in these situations, the
new technology does not encounter resistance from or have to fight
against already existing businesses and entrenched technical
approaches.
Another benefit of strong sponsorship is that it helps protect the
individual risk taker who is willing to take on the entrepreneurial
burden of moving the new technology through the organization. No
matter how beneficial the new technology appears to be, someone
must be willing to sell the effort and make it happen. Schon's (1963)
analysis of successful radical innovation is quite clear. At the outset,
the new technological concept encounters sharp resistance, which is
usually overcome through vigorous promotions by one emerging
champion. What is important to recognize here is that these
champions are typically self-selected; it is extremely difficult to
appoint someone to withstand all of the pressures, hassles, and risks
associated with being an idea chamption and then to expect him or
her to do it excitedly for a long period.
Finally, we also know from research studies that the ultimate use
of a new technology is often not known or may change dramatically
as the technology becomes further developed. The new technology,
moreover, often invades traditional industry by capturing a series of
submarkets, many of which are insulated from competition for some
extended period. The earliest application of the transistor, for
example, was in hearing aids, but its use was not immediately
transferred to the organization's missile divisions. Because of these
more limited niche markets (and consequently, relatively low sales
volume), R&D often concludes that it does not have to work closely
with marketing; nor does it want to subject its technological concept
to the typical market screens of revenue and volume. Such a
conclusion, however, does not help to build the strong harmonious
relationship between marketing and R&D that has been shown to be
so important for successful commercialization of new innovations
(e.g., Souder, 1978). The key to success in these kinds of situations
may be to find a pioneering application where the advantages of
the new capability are so high that it is worth the risks. This would
require the coupling of technical perspective with creative marketing
development to identify such pioneering applications. On this basis,
early involvement of marketing could be very helpful in providing
inputs and market perspective (but not market screens) to the new
technological effort.
4.6. Organizational Rewards

Ultimately, we all know that those activities which are measured
or get rewarded are those which get done. If the managerial and
organizational recommendations and suggestions discussed in this
chapter are to be effectively implemented, then the reward systems
must be consistent and commensurate with the hoped-for behaviors.
One of the most important of these is that research engineers and
scientists must come tl) see that part of their reward system is not just
the generation of publication of new technological concepts und
advances, but that part of their responsibilities is also the successful
transfer of their work. A few high-technology companies we know
have been making such reward systems explicit within their corporate
labs, and although it has taken some time to take hold, it has been
quite effective in moving technology through the development cycle.
It has also resulted in research seeking more joint sponsorship of its
activities, especially with the development divisions-all of which has
helped to strengthen the communication and bridging mechanisms
within the corporation.
Finally, in most areas of day-to-day functioning, productivity
rather than creativity is and should be the principal objective. Even
where innovation and creativity are truly desired and encouraged,
activities that are potentially more creative may be subordinated to

those activities of higher organizational priority or more closely tied
to identified organizational needs. Nevertheless, organizations exhibit
simultaneous demands for routinization and for innovation. And it is
in the balance of these countervailing pressures that one determines
the organization's true climate for managing and encouraging the
introduction of new technological opportunities.
REFERENCES
Allen, T. J., and Marquis, D. G. (1983). Positive and negative biasing sets:
The effect of prior experience on research performance, IEEE Trans.
Eng. Manag. 11, 158-162.
Allen, T. J. (1977). Managing the Flow of Technology, MIT Press,
Cooper, A. c., and Schendel, D. (1976). Strategic responses to technological
threats, Bus. Hor. February, 61-69.
Cohen, H., Keller, S., and Streeter, D. (1979). The transfer of technology
from research to development, Res. Manag. May, 11-17.
Katz, R. (1980). Time and work: Toward an integrative perspective, Res.
Organ. Behav. 2, 81-127.
Katz, R. (1982). The effects of group longevity on project communication
and performance, Admin. Sci. Q. 27, 81-104.
Katz, R., and Allen, T. J. (1982). Investigating the not invented here (NIH)
syndrome, Res. Dev. Manag. 12,7-19.
Katz, R., and Allen, T. J. (1985). Project performance and the locus of
influence in the R&D matrix, Acad. Manag.]. 26, 67-87.
Katz, R., and Tushman, M. (1981). An investigation into the managerial roles
and career paths of gatekeepers and project supervisors in a major R&D
facility, Res. Dev. Manage. 11, 103-110.
Pelz, D. C. and Andrews, F. M. (1966). Scientists in Organizations, Wiley,
New York.
Roberts, E. B. (1974). A simple model of R&D project dynamics, Res. Dev.
Manage. 5, 1-15.
Roberts, E. B. (1979). Stimulating technological innovations: Organizational
approaches, Res. Manage. 22, 26-30.
Roberts, E. B. (1980). New ventures for corporate growth, Harv. Bus. Rev.
July-August, 58, 134-142.
Schon, D. D. (1963). Champions for radical new inventions, Harv. Bus.

Rev. 41, 76-84.
Souder, W. E. (1976). Effectiveness of product development methods, Ind.
Market. Manage. 7,299-307.
Utterback, J. M. (1974). Innovation in industry and the diffusion of
technology, Science 183, 620-626.
Utterback, J. M., and Brown, J. W. (1972). Monitoring for technological
opportunities, Bus. Hor. 15,5-15.
von Hippel, E. (1983). Novel Product Concept from Lead Users: Segmenting
users by Experience, Massachusetts Institute of Technology Working
Paper No. 1476-83.
CHAPTER 12
THE STRATEGIC
MANAGEMENT OF
TECHNOLOGY
Charles S. Skinner
1. TECHNOLOGY IN THE 1970s
In the 1970s, spending on technology declined in the United

States. From 1970 to 1978, U.S. spending on research and develop-
ment (R&D), for example, declined from 2.5% of GNP to 2.1 %,
according to National Science Foundation data. During the same
period, R&D expenditures in West Germany rose from 2.0% to
2.1 % of GNP, and in Japan, from 1.5% to 2.0% of GNP. However,
some technologies such as electronics and data processing continued
to be funded in the United States.
As we enter the 1980s, the impacts of new or rapidly changing
technology are already evident.
Technology is affecting some mature and traditionally low-
technology businesses. (Table I). In the automotive industry, for
example, the wide use of computer-aided design and manufacturing
(including robotics), penetration of microprocessor-based controls
and feature sets, and substitution of man-made materials for natural
ones has affected an industry thought to be staid and static. The same
thing is happening in other mature industries, such as banking,
manufacturing, and publishing.
Charles S. Skinner. Strategic Technology Inc., 24200 Chagrin Boulevard, Cleveland,

Ohio 44122-5598; formerly Vice President, Booz-Allen & Hamilton Inc.
301
302 CHARLES S. SKINNER
TABLE I
New-Technology Applications in Mature Industries
Technology change
Data Telecom-
Market/ industry processmg munications Robotics Materials
Automotive X X X
Banking X X
Manufacturing X X X X
Publishing X X
Consumer goods X
High-technology industries are being revolutionized. One

excellent example is the semiconductor industry. (Figure 1) Between
1975 and 1980, the number of circuit functions per chip on
semi-conductor memories grew from 10,000 to 100,000, while the
price per bit of memory decreased by a factor of 3. The latest
technology, very-large-scale integration, or VLSI, will boost the
SEMICONDUCTOR
MEMORIES
Sales
Bits/Chip (In Billions of Dollars)
,,
1,000,000 I 1.8
1.5
100,000
,
,,
1.2
10,000
,,
BitS/Chip,' Sales 0.9
,
,,
0.6
,
,, ,
1,000
0.3
1
I
1980
I
1990
o
1960 1970
FIGURE 1
Impact of improved technology on high-technology industry.
THE STRATEGIC MANAGEMENT OF TECHNOLOGY 303
number of circuit functions per chip to one million. Additional

improvements in microelectronics capability expected in the later 1980s
or early 1990s will affect markets ranging from computers, produc-
tion machinery, and automotive controls to home appliances, office
equipment, and personal financial services.
Industry competitive positions can be altered significantly, even
to the point some competitors disappear completely. As shown in
Figure 2, in 1975, Sycor was the dominant company in the single
station visual display business. It had the highest sales and return on
sales in the business. But by 1978, Sycor, a poor fifth in a new field
of competitors, was losing money. Microprocessors had replaced
hard-wired logic devices as the cost of microelectronics dropped, a
technology shift that brought computer manufacturers into the
intelligent terminal business and left Sycor out in the cold.
Technology can dislocate entire markets. For example, in the
early 1970s, building automation systems (BAS), which controlled
heating, ventilation, air conditioning, and security in large buildings,
INTELLIGENT TERMINALS
1975 1978
Return on Sales Return on Sales
(Percent) (Percent)
15 30
Sycor.
20 Burroughs.
10
10 Datapoint. Nixdorf.
Incoterm. 0
Philips
•
Delta Data
5~~----~----~--~
-10
-20~
•
Sycor
.Basic Four
______~________~
5 10 25 50 10 100 1000
SALES SALES
(In Millions of Dollars) (In Millions of Dollars)
FIGURE 2
Impact of technology changes on competitive positions.
were hard-wired. But such systems required a number of control

points to operate effectively, which made the systems costly and
impracticable. In the mid-1970s, the sharp reduction in costs and
increase in capabilities of microprocessors facilitated the develop-
ment and introduction of computer-controlled energy management
systems (EMS), which ure much more flexible, decentralized systems
with fewer data control point requirements. (Figure 3).
Technology can affect market-share composition substantially. In
addition to dislocating a market, technology can affect the share of
market held by the major participants. For example, the introduction
of EMS (see Figure 3) in the latter half of the seventies altered the
share of market held by the main participants (Figure 4). Although
the market leaders, Honeywell and Johnson, retained leadership
positions, they held a smaller share of market--40% in 1970 versus
30% in 1979, as shown in Figure 4.
In addition to dealing with these impacts of technology, U.S.
industry is faced with the need to compete in the technology field
with world-scale companies. No longer does General Electric (GE)
compete primarily with domestic concerns like Westinghouse.
COMMERCIAL
ENERGY·CONTROL SYSTEMS
Cost per Point

$2,500
\~:rd'Wired
2,000
,AS
1,500
1,000
500 Program.Controlied EMS
Number of Control
Points Required
FIGURE J
Impact of technology on an entire market.
COMMERCIAL
ENERGY·CONTROL SYSTEMS
(Percent of Total Sales)
100%
I
30
I
I
I f
I
I I
I I 30
I f
I I
I I
I I
Q80"/' /
100% ',' -'
Honeywell 40
FIGURE 4 Johnson 40 "
Share of market by major Other
participants. 1970 1979
Rather, the stiffest competition GE and Westinghouse face at home

and abroad comes from Siemens, Philips, and Hitachi. Detroit's "Big
Three" must square off against Nissan and Toyota, and DuPont must
deal with Hoechst and Bayer.
The results of this marketplace confrontation do not warm the
hearts of American managers. Between 1960 and 1978, for example,
the United States gained position in no industry, and lost position in
12! Japan, on the other hand, increased its position in eight
industries, while losing ground in none. Even the United Kingdom
was able to hold its leadership position in three industries over the
18-year period.
The reasons the United States has faired so poorly in world
technology market come readily to mind.
The high costs of technology development, in general, create a
"natural" barrier to market entry. In addition, the cumulative
expenditures on R&D, especially if concentrated in selected product
lines or on key technologies, builds the barrier even higher. This
situation already exists in such industries as chemicals, aerospace,
automotive, electrical equipment, machinery, steel, and
pharmaceuticals.
The key to successful market entry, then, is economies of scale,

possible only through volume production and sales. But because the
United States is blessed with substantial, diversified, und dynamic
domestic markets in general, we have been slower than our com-
petitors (e.g., Japan, West Germany) to seek out and develop
sufficiently extensive export markets for our cornucopia of technical
goods and services.
2. IMPACT OF TECHNOLOGY ON BUSINESS IN THE

19805
In a survey we conducted in 1979 among 2,000 officers of the

Fortune 1,000 group of companies, the respondents listed produc-
tivity, energy costs, new product development, and rapidly changing
technology as four of their top eight concerns for the 1980s (Table
II). Japanese and European managers, sampled on a more limited
basis, generally agreed with this view of technology.
By way of comparison, the results of a 1981 survey of 800 senior
managers show only slight shifts in concerns. Changing technology
and rising energy prices remain major concerns, joined by worries
about cost of capital and government regulation (Table II).1
Despite this continued awareness of the importance of technol-
ogy to business operations, respondents in the more recent survey
TABLE II
Top Concerns of CEOs
1979 1981
Costs of capital X
Increasing energy costs x X
Productivity X
Government regulatory policies X
Changing technology X X
New product development X
Foreign competition
Foreign government economic
and trade policies
view technology as difficult to manage. For example, 86% of the

respondents indicated that a lack of an analytical approach to
ihtegrating technology and planning is a barrier to managing technol-
ogy. And 80% indicated a limited involvement of technology
managers in the planning process as a drawback. Moreover, only
37% of managers sampled think their companies effectively forecast
technological change in this country, and only 13% believe their
organizations do effective forecasting of foreign technological
change.
Similarly, survey respondents felt technology-based functions
such as research and development are of less importance to overall
corporate performance than line-operating functions (e.g., manufac-
turing, marketing, and sales) or planning functions. And the same
managers believe technology-based functions are underfunded, prob-
ably the result of short-term-profit orientation of managers and
external as well as internal pressures for near-term financial perfor-
mance. Specifically, R&D and new product development are per-
ceived to be underfunded by half of all respondents, while only a
quarter believe corporate planning is underfunded (Figure 5). At
other points in the spectrum, some line functions like marketing and
(PERCENT OF RESPONSE)
OVERFUNDED UNDERFUNDED
.•
20 15 10 5 0 5 10 15 20 25 30 35 40 45 50
I , I , I I I I I I I I I I I
Research and development
---
•
New product development
•
Manufacturing and engineering
Production or manufacturing
Corporate planning
Internal information systems
Marketing/sales
Accounting and finance
FIGURE 5
Adequacy of investment funding, by function.
J08 CHARLES S. SKINNER
sales were viewed as appropriately funded, while finance and

accounting may be a candidate for cost reduction.
These somewhat pessimistic views do not, however, dilute
management's appreciation and understanding of technology's im-
portance in the decade ahead. Fully 80% of all respondents indicated
changing technology will have a moderate or extensive impact on
their companies' business operations. When asked to identify factors
that would enhance technological effectiveness, managers listed five
factors (Table III). In order of perceived importance, they are
• Greater top-management commitment to longer-term results;

• Improved long-range technology planning;
• Higher R&D expenditures;
• Integration of technology into business decisions;
• Tighter coupling of technology and strategic planning.
TABLE III
Enhancers of Technological Effectiveness
Most effective
Longer term focus
Management more involved in long-range technology planning
Greater R&D spending
Integration of technology into business decisions
Coupling technology with strategy
Less effective
More recognition for technical people
Incentive compensation for technical people
Dual career paths
Use of outside technical specialists on a project basis
University involvement
Promotion of technologists to senior management
Least effective
Better allocation of R&D expenditures
Greater job security for technical people
Training for technical people
Technical advisory boards
TABLE IV
Differences in Technology Management Approaches
Leaders Followers
Perspective Long Short

Most influential investment Technological Costs
factor change
CEO/COO involvement Deep/long Shallow/short
Technology management/business Well integrated Isolated
strategy
Moreover, some companies have made real progress in integrat-

ing technology with long-range planning and strategy. For example,
12 technology leaders included in a Booz-Allen study on the
management of technology,2 sponsored by General Electric in the
late 1970s, generally have adopted a longer-term perspective, and
view technological change as the most influential factor for long-term
success. In these large, sophisticated companies, top-level executives
are involved in the management of technology, and technology is
integrated in the business strategies (Table IV).
With the exception of these leaders, however, management views
technology as important but complex and difficult to manage.
Technology managers are not involved in business planning, and
senior managers are not involved in the development of a technology
strategy. Clearly, there is a gap between the perceived importance of
technology in business and the attention given it by managers.
The first steps to correcting these problems are to view
technology as a strategic resource; elevate it to a level of importance
paralleling that of business planning; and develop a formal, analytical
approach to integrating technology in the planning process.
3. BOOZ-ALLEN'S VIEW OF TECHNOLOGY

MANAGEMENT
Contrary to traditional views, the path and timing of technology

development are, in Booz-Allen's view, predictable. In a recent study,
we found the pattern of market proliferation and timing were very

similar for dozens of inventions. As shown in Figure 6, and using
xerography and integrated circuits as examples, technology develop-
ment moves through three phases: watch, get ready, und go.
We also differ from the traditional view of how available
resources should be allocated to technology. The traditional view is
that technology strategy and investment are determined separately
from the overall business strategy. In contrast, we believe, first, that
the technology can be planned as a capital asset; and second, that the
importance of the individual technology to both the technology
strategy and the business strategy determine the resource allocations.
The challenge is to ensure consistency between the amount of
scarce capital and human resources invested in various technologies,
and the business objectives to be achieved. Too often in our work for
technology-based companies, we find a mismatch between strategic
objectives and technology investments. Insufficient, excessive, and
improper investment allocations are common.
For example, companies fail to invest adequately to maintain
their technology leadership positions after establishing them. They
stop investment short of developing a proprietary technology. They
do not spend the time and money to monitor the competition,
particularly on a worldwide basis. And sometimes they do not invest
enough to change their own competitive position when they have a
market opportunity to do so.
Xerography
• Proliferation
GO • Acceptance
GET • Commercial
READY introduction
• Field trial
WATCH • Verification
• Origination
FIGURE 6
Pattern of technology development.
On the other hand, some companies spend excessive amounts on

developing new products for maturing businesses-businesses their
competitors are harvesting, or businesses with low opportunity for
high yield. This occurs most frequently in companies in which
technology investments are based on, or influenced by, a percent-of-
sales measure.
Improper allocation of available financial resources is also
common. Among the examples are: reducing expenditures on pro-
grams for promising new products when the company should be
spending to secure a market position, investing in technologies in
which a competitive advantage is impossible, spending for defensive
programs, and spending on programs at odds with the corporate
portfolio strategy. In each case, the net result can be an eroding
competitive position.
4. BOOl-ALLEN'S APPROACH TO TECHNOLOGY

STRATEGY
Correcting the imbalances of technology investment and strate-

gic objectives requires integrating technology investment strategy
with business strategy. Booz-Allen's approach to meeting the need
involves:
Developing the technology portfolio. Development of a technol-
ogy portfolio is predicated on the belief that technology change and
competitive positions can be analyzed and measured in terms of
market attractiveness versus relative competitive position, and that
the position of a technology is critical to overall competitiveness.
Matching business and technology portfolios to identify oppor-
tunities for improving market position.
Setting technology investment priorities, and thus an investment
strategy for funding and allocating resources consistent with the
business strategy.
4.1. Developing the Technology Portfolio

In developing a technology portfolio, first, businesses sensitive
to technology change are identified by segmenting each business into
major products and processes, and identifying the principal technolo-

gies that contribute to the competitive position of each product und
process. Next, past and current technology investments are exam-
ined, issues on the appropriate and necessary level and rate of
technology investment are identified, and questions of additional
technologies and corresponding investments required to achieve
corporate business objectives are raised.
The outcome of such an examination is a definition of the
company's internal technology situation in terms of its technological
capabilities, the extent to which value is added by its technology, and
the importance of its technologies to the company's overall success.
Then the competitive environment is scanned to gain an
understanding of the investment levels and investment patterns of
competitors for both the product and process side of each technol-
ogy. Competitors' technology-investment concentrations and their
positions in technologies critical to your success need to be deter-
mined as well.
The result of these critical internal and external analyses is a clear
understanding of a company's competitive position in each of its
technologies, and the relative importance of each technology in the
marketplace. With this information, a technology portfolio can be
plotted on the same four-quadrant grid used for business strategy
analysis (Figure 7). The use of this format aids in understanding the
integration of the technology strategy with the business plan.
The vertical axis, technology importance, is usually defined as
the value from investing in a technology, the potential for changing
High
BET DRAW
Technology
Importance
CASHIN FOLD
Low
High .. Low
Relative Technology FIGURE 7
Position The technology portfolio.
that technology's competitive position, the market-growth potential

of all applications for the technology, and the importance of having a
superior technology to compete in the markets over the long term.
The horizontal axis, relative technology position, represents a
technology's current and future position and expected future
development. Some of the quantitative criteria used to position each
technology are amount of investment-R&D and capital investment,
for example-human resource allocations, new product development
success, and manufacturing cost positions.
The four quadrants are labeled "bet," "draw," "cash in," and
"fold." If, after analysis, technologies land in the "bet" quadrant
(Figure 8), the company is in an excellent position technologically-
for example, it is in business segments in which those technologies
are important. In this situation, the company should keep its
investments in these technologies high to sustain competitive advan-
tage, and allocate the necessary top-level manpower to these winners.
Technologies in the "draw" quadrant are borderline cases; that
is, they are important, but are in marginal competitive positions. In
this situation, one of two choices is wise (Figure 9): bet against the
competition by investing to attain a leadership position, or "fold" the
technology and redirect investments to more lucrative areas.
Technologies in the "cash in" quadrant indicate the company
enjoys a strong position technologically but the technologies them-
selves are not very important in the marketplace. This situation
occurs most frequently in a rapidly changing industry, such as
electronics or engineered plastics, in which existing technologies are
High
BET-
•
DRAW
Technology
Importance
CASHIN FOLD
Low
FIGURE 8 High .. Low
The technology portfolio. (Size of circle Relative Technology
reflects relative size of investment.) Position
High
BET .... f-O DRAW

Technology
Importance
•
CASH IN FOLD
Low
FIGURE 9
High ... Low
The technology portfolio. (Size of circle
Relative Technology
Position reflects relative size of investment.)
quickly supplanted by new technologies. The approach needed is to

reduce investment selectively (for example, selling off the technology
through licensing agreements or franchises) and redeploy funds to
more promising technologies.
Technologies in the "fold" quadrant are twice-poor: they are
not important, and the company does not enjoy a strong competitive
position. Depending on other criteria, the decision here becomes
obvious-and the quicker the better!
4.2. Matching Technology and Business Portfolios
The business portfolio, which is product-oriented and measures a

company's product lines in terms of market position and market
importance, and the technology portfolio, which is technology-based
and defines the company's product and process technologies in terms
of competitive position in the marketplace and the importance of
each technology to the company's basic business, must work
together. Although they provide fundamentally different views of a
company, those views must be essentially congruent if a sustainable
competitive advantage in a technology-sensitive business is to be
achieved. For example, as shown in Figure 10. Product A is in the
"bet" quadrant of the business portfolio: that is, the company holds
a strong competitive position in a highly attractive maket. But the
company is not comparably strong in Technologies A that drive
Product A. Thus, if Product A's competitive position is to be
sustained, a decision needs to be made about upgrading technology
investments in the short term.
BUSINESS TECHNOLOGY
° 0 t°
•°
A
Q)
(J A
c:
!!! 0 0 •
o
0-
0
E
..
~Competltlve-
0°
Position
10 ..
0°
_____ Poslhon---
FIGURE 10
Portfolios. (Technologies aggregated to represent technological context of businesses
presented in business portfolio.)
A technology portfolio, viewed in the context of a business

portfolio, serves a number of purposes:
• It establishes a common planning base for all technologies.

• It provides a summary of a company's technology position, an
understanding of corporate technological focus and a method
for synchronizing the timing of corporate technology in-
vestments with the business plan.
• It identifies positions of strength to be leveraged. It also
identifies technology requirements to be strengthened or
acquired to achieved corporate business objectives and com-
petitive advantage.
• It provides a basis for focusing on high-potential new business
opportunities that could use additional technology investment.
4.3. Setting Technology Investment Priorities

Arranging technologies in terms of quadrant position along an
axis of relative technology expenditures is a convenient way of
summarizing a company's technology strategy options and initiatives.
(Figure 11) For example:
• For a technology in the "bet" stratum, expenditures should be

BET
DRAW
CASHIN
FOLD
-Relative • FIGURE 11
Expenditures Technology investment priorities.
high, relative to the leader. If they are not, management

should consider increasing investments in this technology .
• For technologies in the "draw" area, management should
either increase expenditures to move the technology to the
"bet" area, pull back (for example, cash in or fold), or play
one more round (draw). In the longer term, however, the third
option is highly undesirable because it drains away needed
capital from other, more promising alternatives .
• For technologies in the "cash-in" or "fold" strata, man-
agement should level out or reduce investments so that
available resources can be allocated to technologies with
higher leverage.
Technology strategy, then, is built on a penetrating analysis of a

company's technology strengths and weaknesses, relative to the
importance of these technologies to its businesses. Together with the
business strategy, the technology strategy defines how a company
can most effectively invest the company's technology resources to
achieve sustainable competitive advantage.
NOTES
1. Although "productivity" was not included in the list of possible choices

in the 1981 survey, the results of other research conducted recently by
Booz-Allen suggest it remains a major concern of senior management.
2. Of the twelve, eight were U.S. firms, three were Japanese, and one was
European. Their total gross sales exceeded $90 billion (in 1980 dollars),
and their total expenditures on R&D exceeded $3.5 billion.
SUBJECT INDEX
A management, 200, 201, 203, 204, 206, Automotive industry, 146, 170,301
208, i12, 219 robot use, 88, 89
X-efficiency, 211 Automotive manufacturing, 2, 57-71
Absenteeism, 217, 218
Accessories costs Basic metal cost index, 101, 103
insensate robots, 18-20 Basic oxygen furnace, 141, 142
Accounting Batch production
productivity management, 4 impact of flexible manufacturing sys-
Accounting systems tems, 94-107
manufacturing, 70 Bayer, 305
Acquisition Bell Laboratories, 134
corporate priority, 261 Bendix, 21
Advanced technical groups Bessemer converter, 142
within development, 294 BMCI, 101, 103
Air-cooled aircraft engines, 139 Booz-Allen and Hamilton, 154
Air Products 145, 146, 293 Bounded rationality, 207, 209
Aircraft engines, 135, 139, 139, 140 Building automation systems" 303, 304
Ajax process, 142 Business
American Machinist Inventory, 75 impact of technology, 306--309
American Robot Corporation, 22 Business strategy
AMF Versatran, 23 technology strategy and, 310, 314, 315
Architectural design
manufacturing control systems, 166 CAD/CAE, '191,194
Artificial intelligence, 2, 10, 48 CAD/CAM, 3, 4, 161
flexible manufacturing systems, 49, 50 CAM, 268, 269
ASEA 21 Capacity utilization
Assembly productivity measurement, 240
flexible manufacturing systems for, 47, Capital costs
48 effects of increased output, 96
sensate robots, 13 robot installation, 93, 94
Automation, 268 Capitalist firms
batch production, 99 classical, 211, 212
management attitudes to, 189-195 Career patterns
stamping operations, 62-64 under Z management, 201
vehicle assembly, 59-62 Carnegie-Mellon Robotics Institute, 47, 48
Automatix, 31 Carnegie-Mellon University, 24
317
318 SUBJECT INDEX
Cells Consumer electronics industry, 170, 171

manufacturing, 37, 39, 46 Contacts
Celluloid roll film, 128, 136 informal, value of, 279, 280
Change Continuous casting, 143
discontinuous, 115, 117, 118-122 Contractural structure
repeated waves of, 131, 132 capitalist firms, 211, 212
Chassis shops Control decisions, 165-167
vehicle assembly, 60-62 Control systems
Cincinnati Milacron, 20, 23 flexible manufacturing systems, 35-37
Closed-loop joint control, 15 Cooperation
end effector positioning, 26 employees, 209, 210 '
Cluster analysis nonreciprocal, 222, 223
MPPS, 172, 174 Corfam, 293
Cognitive processes Corporate culture
restricted, 283-285 Z management, 201
Colburn, 142 Corporate sponsorship
Collective decision making of new technology, 296-298
Z management, 201 Corporate strategy
Color television, 148 Z management and., 219, 220
Committment Corporate venture strategies
employees, 205 for new technical approaches, 291, 292
Communication behavior Cost leadership, 156
engineering professionals, 282, 283 Cost/performance analysis
Communications MPPS, 174, 175
technical, 278-280 Cost reduction
Companies effects of new technology, 131, 132
Japanese, 201, 202 Cost structure
Competition manufacturers, 271, 272
international, 304, 305 Costs
Competitive positions implicit psychological contracts and,
effects of technology changes, 303, 312 206-210
Competitive pricing, 175 insensate robots, 18-20
Competitive priorities of new technology, 293
Manufacturing Futures Survey, 261-263 of noncooperation, 209, 210
Competitive strategy, 3, 113, 126-133, production, 177
153 robot installation, 86, 87
Computer-aided design, 191, 194 robot systems, 50
Computer-aided engineering, 191, 194 sensate robots, 90
Computer-aided manufacturing, 161, 162 working monitoring, 216, 217
management attitudes to, 189-195 Custom producers, 97
Computer assisted manufacturing, 268, 269 Cylinder-blowing process, 142
Computer-controlled robots, 23-25
Computer integrated manufacturing, 57,
68,69,70 Decision making
Computer languages collective, 201
rule-based, 48, 49 Z management, 219
Conflict resolution Demand
Z management, 219 effects of discontinuous change, 148
Consensus formation price elasticity, 90-94
decision making, 201 Department of Defense, 122
SUBJECT INDEX 319
Design Energy management systems

new products, 115 computer controlled, 304
Diesel electric locomotives, 122, 128 Engine blocks
Direct labor aluminum, 129
costs, 70 Engineering professionals
motivation programs, 268 communication behavior, 282, 283
reduction, 190, 191 value of informal contracts, 279, 280
savings, 82-90 Environment
Direct-drive manipulators structuring, for robots, 17, 18
robots, 24, 25 Establishments, larger
Discontinuous change, 115, 117 robot concentration in, 88-90
pattems of, 118-121 Evaluation
Discounting managerial performance, 239
future utilities, 214, 215 Evolutionary change
Divestiture in technology, 113, 114
corporate priority, 261 Expectancy theory, 223
Dodge, 146 External sensing, 48
DuPont, 293, 305 robots, 11-14, 25-31
Eastman, George, 122 Facility decisions, 164, 165

Eastman Kodak, 128 Fanuc Ltd, 43, 45, 46
Economics Feasibility
flexible automation systems, 189-195 of new technical concepts, 294
manufacturing strategy and, 233-236 Field effect transistors, 135
Economies of scale, 158, 159, 160, 161, Firms
175 organizational structure, 167, 168
Economies of scope, 159, 160, 161 response to new products, 125-129
Economies of speed, 158 Flat glass, 146
Eikonix Corporation, 17 Flexibility
Electric illumination, 129 in manufacturing, 162, 164, 166
Electric razors, 136 Flexible machining, 57
Electric typewriters, 122, 140 Flexible manufacturing systems, 3, 9, 10,
Electronic calculators, 135, 136 33-39,69, 161
Electronics assembly applications other than machining, 47--49
flexible manufacturing systems for, 47, costs, 40--44
48 economics, 189-195
Employees effects on manufacturing costs and
relations among, 201, 208-210 employment, 74
Employer--employee relations, 222 floor space requirements, 51
implicit psychological contract, 204--210 for machining of parts, 39--46
Japanese, 202 impact of batch production, 94-107
Z management, 219 multiple, 47
Employment sensor-based and AI systems, 49, 50
effects of robotic manipulators, 76-94 Float process, 146
impacts of new manufacturing technolo- Floor space requirements
gies,2 flexible manufacturing systems, 51
lifetime, 200 Focus
End effectors in manufacturing, 159
positioning, 14-16, 26 manufacturing control systems, 166
robots, 10 Ford Model A, 137
320 SUBJECT INDEX
Ford Model T, 137 Information

Ford Motor Company, 2, 57, 58-67, 122~ for research and development, 278-280
137-139, 146, 147 Information management
Forging of factory automation, 194
flexible manufacturing systems for, 47, Information processing
48 selecti ve, 282, 283
Fourcault, 142 sensory, 29, 30
Functional areas Innovation, 3, 4
emphases, 271 corporate sponsorship, 296-298
organization structures, 287-292
Gas illumination, 129 organizational controls for, 292-299
General Electric Co., 3, 21, 42, 44, 46, patterns of, 113-121
147,191,293,294,304 radical
Laundry and Dishwasher Products Divi- see radical innovation
sion, 192, 193 reversal, 142, 143
Transportation Systems Division, 193, Insensate robots, 12-14
194 costs, 18-20
General Motors, 122, 146 job displacement, 92, 93
General Purpose Plant Strategy, 165 limitations, 14-21
Glassmaking, 142, 146 payback periods, 85
Global Manufacturing Futures Study 1983, replacing production workers, 76, 77
259--272 retrofitting, 84
GM-Fanuc,21 Inspection
sensate robots, 13
Hardware costs Installation costs
flexible manufacturing systems, 40-44 insensate robots, 18-20
Hierarchical control systems robots, 85, 86
flexible manufacturing systems, 35-37 Integrated circuits, 140, 148
Hitachi, 305 Interaction
Hoechst, 305 selective, 283
Honeywell, 304 Internal sensing
Human bridges robots, 11-14
research and development transfer, 290 Interpretation
Human sensing of sensory signals, 27, 28
compared with machine sensing, 29, 30 Inventory control signals, 165, 267, 268
Inventory policy
IBM, 21, 47, 122, 147 just-in-time, 57, 68, 165
Ice-making, 127, 128, 144 Investment
mechanical, 121 in flexible manufacturing systems, 3,
Ice refrigerators, 136 190
Iceboxes, 127, 128 in technology, 310, 311
Imperfect competition, 234 transaction specific, 231-233
Implicit contract, 203, 204
Implicit psychological contract, 203, 204- Japanese management
210,220,230,231,232 see Z management
Indirect labor productivity, 262, 263, 264 Jet engines, 135
Industrial diamonds, 144 Johnson, 304
Industrial gases Joint motions
on-site production, 145, 146 insensate robots, 15
Industrial robotics, 2 Just-in-time, 57, 68, 165
SUBJECT INDEX 321
Kaldo converter, 142 Liquid oxygen

Kearney & Trecker Corporation, 37-39, on-site production, 145
40-42,44 Load disturbances
Kinematic arm solution joints, 15
joint motions, 15 Lubber cylinder-blowing machine, 142
Machine Intelligence Corporation, 31

Labor availability Machine sensing
Japan, 268 compared wtih human sensing, 29, 30
Labor contracts, 208 fundamental problem of, 27-31
Japanese, 222 Machine utilization
Labor costs effects of increased output, 97
effects of increased output, 96 Machine vision
justification for installing robots, 82, 83 commercial applications, 31-33
Labor force uses, 30, 31
attrition, 93 Machining
Labor hours flexible manufacturing systems for, 39-
productivity measurement, 246, 247, 46
248 improved processes, 68, 69
Labor-management relations sensate robots, 13
changing, 266 Maintenance
Labor market improving, 266
Japanese, 202 Management
transaction costs, 207, 208, 222 Japanese, 3, 4, 199, 218
Labor motivation programs, 268 see also Z management
Labor productivity, 203, 263, 263, 264 Management attitudes
Japanese, 200 to automation, 189-195
Labor turnover, 217, 28, 224 Management control, 239-242
Labor usage rate, 243, 244 defined, 240
Labor variance productivity management, 4
productivity measurement, 240 Management hierarchies
Laundry and Dishwasher Products Division vertically integrated, 158
General Electric, 192, 193 Management information systems, 267
Lawrance, 139 Management style, 271
Lead-through programming, 24 Management techniques
end effector positioning, 14 Japanese, 3, 4
Level I robots, 12-14 Managers
costs, 18-20 background, 271
job displacement, 76, 77, 92, 93 evaluation, 239
limitations, 14-21 Manipulation
payback periods, 85 robot applications, 12, 13
retrofitting, 84 Manipulators
Level II robots, 12 robots, 24, 25
costs, 90 Manual typewriters, 140
replacing production workers, 76, 77 Manufacturers
Lifetime employment, 200, 215 improving effectiveness, 265, 266
Z management, 206 major concerns, 263-265
Light manufacturing organization structure, 270
high-precision robots for, 21-23 product concentration, 271
liquid-cooled aircraft engines, 139 product management, 272
322 SUBJECT INDEX
Manufacturers (cont.) Materials usage per unit output, 242

profit margins, 270 Mazak Corporation, 47
programs for the future, 267-269 Mechanical drawing
Manufacturing sheet glass, 142
accounting systems, 70 Mechanical ice-making, 144
computer assisted, 3, 268, 269 Mechanical refrigerators, 136
process optimization, 179-192 Meetings
Manufacturing cells, 37, 39, 46 problem solving, 290
Manufacturing flexibility, 162, 164 Metalworking industries, 2, 74-76
Manufacturing Futures Survey, 259-272 output measurement, 94, 95
Manuacturing mission, 156, 168, 170 Microeconomics
and control systems, 166 and Z management, 229-233
Manufacturing planning support systems, Mobility
154,171,172 within organizations, 286, 287
functions and design, 172-184 Monitoring costs
implementation, 183-185 shirking behavior, 216, 217
Manufacturing strategies, 3, 4, 153-155 Monsanto Company, 287
decisions and performance, 155-168 Motivation
institutional economics and, 233-236 workers, 203, 204
paradigm, 168-172 workers, 210-218
Manufacturing systems, 9 MPPS, 154, 171, 172
performance prediction, 183 functions and design, 172-184
types of, 33-35 implementation, 183-185
Manufacturing technology
new, 1,2 NASA, 145
Market dislocation Night shifts
effect of technology, 303, 304 unmanned, 45
Market entry Nissan,305
effects of technology development, 305, North American Automotive Operations
306 Ford Motor Co., 58-67
Market Plant Strategy, 164, 165 Not invented here syndrome, 285, 286
Market response model Numerically-controlled machine tools
MPPS, 182, 183 contrasted with robots, 10, II
Market segmentation, 156 Nylon, 140
Market share
effects of new technology, 304-306 Off-line programming
Marketing robots, 23, 24
links with research and development, Office automation, 269
293, 294 Ohain, von, 135
Markets On-line programming
new products, 121-123 robots, 24
Mass producers, 97 Open-die forging
Mass production, 97, 98 flexible manufacturing systems for, 47,
Material costs 48
Japanese producers, 261 Open hearth furnaces, 142
Material handling Open-loop joint control, 16
random bypassing capability, 35 Operating costs
Material requirements planning, 166, 167 flexible manufacturing systems, 40-44
Materials Operating strategies
productivity measurement, 247, 248 automotive industry, 67-70
SUBJECT INDEX 323
Opportunism, 207, 209 Philips, 305

Organization decisions, 167-168 Photographic film, 122, 128
Organizational changes Pilkington Glass Ltd., 146
research and development transfer, 291 PILOT, ISS, 175-179, 183-185
Organizational control Piston engines, 135
implicit psychological contract, 206 Pittsburgh Plate Glass Company, 142
Organizational controls Planning
for innovation, 292-299 integration of technology with, 309
Organizational goals organizations, 275
Z workers, 215, 216 Plant
Organizational structures capacity, 246
for innovation, 287-292 configuration, 178
managers, 270 location, 164, 165
Organizations size, 164, 165
effects of new technology,S, 281, 282 reconditioning, 266, 268
planning and innovation, 275 Plate glass, 122
short and long term goals, 275-280 Population forecasts
uncertainty in, 280 robots, 77, 84
Orientation devices Positioning systems
robots, 17 end effectors, 14--16
Output Pounds of metal processed as output meas-
metal working industry, 10 I-I 03 ure, 94, 95
Output:input partial productivity measure, Prab, 21, 23
242-244 Pratt & Whitney, 140
Output measurement Price
metalworking industry, 94, 95 priority, 262
Overhead costs, 263, 264 Price elasticity
and demand for robots, 90-94
Pricing
Paint shops competitive, 175
vehicle assembly, 60, 61. 62 PRIME, ISS, 179-182, 183-185
Partial productivity measures, 240, 242, Principal-agent relationship
243 employer-employee, 205, 206, 211, 222
Parts assembly Priorities
high-precision robots for, 21-23 manufacturing control systems, 166
Parts feeding Problem solving
robots, 17 homogeneity, 284, 285
Parts machining meetings, 290
flexible manufacturing systems for, 39- rigidity in, 280, 282
46 Procedural bridges
Payback periods research and development transfer, 289,
robot installation, 85-88 290
People Procedural computer languages, 48
movement within organizations, 286, Process
287 changes, effects on product, 143-146
Performance decisions, 159-164
incentives, 4 design, 162
managers, 239 discontinuities, 130--132, 141-143
new products, liS, 116 new, 268, 269
Personal and situational uncertainty, 280 optimization, 179-182
324 SUBJECT INDEX
Process (cont.) Productivity

radical innovation, 114, 133-137, 141- defined, 241, 242
143 effects of product quality, 191, 192
Process focussed firms, 167 Japanese, 200, 218
Process Plant Strategy, 165 relationship with X-efficiency, 202, 203
Process-product discontinuities, 143-146 Productivity index, 240
Process technology Productivity management, 3-5
Japanese, 263 Productivity measurement, 239-342
new, 268 benchmarks, 241
Process tools consistency with profit maximization,
stand-alone, 37, 38 242-245
Product framework for, 245-253
concentration, 271 surrogates, 240
decisions, 156-158 Products
demand, 148 new, 268, 269
design, 157, 158, 162 revolutionary, 133-137
development, 307 Profit margins
differentiation, 156 manufacturers, 270
new, 268, 269 Profit maximization
performance, 115, 116,262 consistency with productivity measure-
proliferation, 159 ment, 242-245
quality, 191, 192,262,263 Profit variance analysis, 251, 252
radical innovation, 114, 130-132, 137- Programming
141 robots, 51
redesign, 266, 268 PUMA, 22, 23
revolutionary, 133-137
Product demand attributes, 157 Q-BOF, 142
Product focussed firms, 167 Quality, 191, 192,262,263
Product management Quality circles, 268, 269
manufacturers, 27
Product manufacturing attributes, 157 Radical innovation, 114
Product Plant Strategy, 164 model of, 121-133
Product-Process discontinuities, 2, 3, 133- response, 123-133
137 Random bypassing capability
Production material handling, 35
modes of, 97, 98 Rayon, 122
Production control systems, 267, 268 Rayon, 140
Production cost standards RCA, 148
productivity measurement, 240 Refrigerators
Production costs, 177 electric, 127, 128
Production/distribution model, 183-185 Repeatability
MPPS, 175-179 robots, 50
Production process model, 183-185 Research and development
MPPS, 179-182 communication behavior in, 282, 283
Production scheduling isolation, 282, 283
automotive manufacturing, 64-67 not invented here syndrome, 285, 286
Production workers place in the organization, 275-278
reabsorption, 93 problem solving strategies, 280-282
replacement by robots, 73, 74, 76-94 support for new technology, 294, 295
SUBJECT INDEX 325
Research and development (cont.) Semiflexible transfer lines, 39

transfer of results, 288-291 manufacturing systems, 34, 35
underfunding, 307 Sensate robots, 12
Responsibility costs, 90
manufacturing control systems, 166 replacing production workers, 76, 77
Results Sensing
joint appraisal, 289, 290 robot applications, 12, 13
Risk aversion Sensor-based systems
Z workers, 215 flexible manufacturing systems, 2, 10,
Robot Institute of America, 88 49,50
Robot systems Sensory information processing, 9, 29, 30
costs, 50 robots, 25-31
Robotic manipulators, 2, 9, 10, II Service
effects on manufacturing costs and priority, 262
employment, 76-94 Servo control
retrofitting, 73 robots, 10, 11, 51
Robotics, 161 Sharp Co., 135
Robots Sheet glass
batch production, 99 mechanical drawing, 142
communication, 24 Shirking, 212
computer-controlled, 23-25 Z management, 216, 217
external sensing, 25-31 Siemens, 305
high-precision, 21-23 Signals
internal and external sensing, 11-14 not accessible to humans, 29
payback periods, 82-90 Silicon growth junction transistors, 134
population forecasts, 77, 84 Small parts assembly
programming, 51 high-precision robots for, 21-23
repeatability, 50 Software development
replacing production workers, 73, 74 for MPPS, 154, 155
use in larger establishments, 87-90 Specialization
Rolls Royce, 140 restricted perspective, 283-285
Rotor process, 142 Specialized transfer lines, 38
Rubies manufacturing systems, 33, 34
artificial, 144 Speed economies, 158
Rule-based computer languages, 48-49 SRI system
machine vision, 31
Safety razors, 136 Stamping operations
Sapphires Ford,62-64
artificial, 144 Stand-alone process tools, 37, 38
Scale economies, 158, 159, 160, 161, 175 Statistical process control, 57, 68
Scheduling instability Steam locomotives, 128
automotive manufacturing, 65-67 Steel industry, 143
Scheinman, Victor, 23 Steelmaking, 141, 142
Scope economies, 159, 160, 161 Strategic decisions
Scoville Corporation, 128 FMS investment as, 190
Segmentation analysis Strategic management
MPPS, 172, 174 of technology, 301-316
Seiko Instruments, 21, 22 Strategic priorities
Semiconductor industry, 302, 303 Manufacturing Futures Survey, 261-263
326 SUBJECT INDEX
Str.ategy Time preference, 215

integration of technology with, 309 Japanese workers, 223, 224
Strategy relevant resources, 234, 235, 236 Tool positioning
Structuring sensate robots, 13
robot environment, 17, 18 Tools
Sumlock Co., 135 stand-alone, 37, 38
Supervisor training, 268 Toyota, 305
Surrogates Training
productivity measurement, 240 for new operating strategies, 70
Swan Lamp Company, 122 Transaction cost economics, 203, 207
Sycor,303 Transaction costs, 200, 221, 230, 231
Synthetic gems, 144, 145 labor market, 222
Transaction costs economics, 230
Task positions Transaction specific investments, 231-233,
industrial robots, 16, 17 236
Technical communications, 278-280 Transfer
Technical feasibility, 294 research and development results, 288-
Technical understanding 291
of new technology, 292-294 Transfer groups, 291
Technological change, 113-121 Transistors, 134, 135, 140, 292, 293
continuous, 296, 297 Translation
Technologies sensing as, 27-28
limitations, 292, 293 Transport
Technology robot applications, 12, 13
assessment of new, 252, 253 Trim shops
defense of old, 282 vehicle assembly, 60-62
diffusion as long term process, 295, 296 Turbofan aircraft engines, .139, 140
effects on high technology industries, Turbojet aircraft engines, 135
302, 303 Typewriters, 140
effects on mature industries, 301
effects on organizations, 281, 282 Uncertainty
impact on business, 306--309 motivation to reduce, 280
invasion of new, 114-121 Unimate,23
investment, 310, 311, 315, 316 Unimation, 16, 20, 22, 47
manufacturing control systems, 166 Unions
modernizing, 129, 130 Japanese, 202
perceived as difficult to manage, 306, Unit costs, 77
307 determination, 175
strategic management, 301-316 effects of robots, 73, 74
transfer, 288-291 flexible manufacturing systems, 74, 104-
underfunding, 307, 308 146
Technology transfer groups, 291 impacts of new manufacturing technolo-
Texas Instruments, 148 gies,2
Theory X, 206 invading product, 117, 118
Theory Y, 206, 221 metalworking industry, 101, 102
Theory Z, 206 new products, 123, 124
see also Z management Utility-effort relation, 210-218
Thin-wall casting, 129
Three M Corporation, 287 Vacuum tubes, 134
SUBJECT INDEX 321
Value added Worker motivation

effects of increased output, 96 A and Z management, 203, 204
per employee, 271 Z management, 210-218
Value added per labor hour, 242 Workers
Value analysis, 268 broader range of tasks, 266
programs, 266 membership behavior, 217, 218
Vehicle assembly monitoring, 212
Ford, 59--62 Wright Co., 139
Venture capital investment, 291 Wyle Co., 135
Vertical integration
corporate priority, 261 X-efficiency, 200, 210, 211, 220
management hierarchies, 158 Japanese firms, 199
VICARM,23 relationship with productivity, 202, 203
Vision systems, 27 xerography, 136
Xerox, 288
Walk-through programming, 24
end effector positioning, 14
Wasada University, 51 Yamazaki, 47
Westinghouse, 304 Yamazaki Machinery Works, 43, 44, 46
Wet copying, 136
Whirlpool, 293 Z management, 208, 218-220, 221, 224
White collar productivity, 263, 264 applicability, 231-233
White-Sundstrand, 39 characteristics, 200-202
Whittle, Sir Frank, 135 lifetime employment, 206
Work force microeconomics and, 229-233
aging Japanese, 264, 268 transferability, 219
Work patterns worker motivation, 203-204, 210-218
commitment to, 281 zone melting technique, 134
AUTHOR INDEX
Abernathy, J. W., 138 Caves, R., 233, 235 Fitzroy, F. R., 216, 217,
Abernathy, W. J., 113, 114, Chandler, A. D., 158 224
119, 120,221 Charnes, A., 245 Frank, S., 142
Agin, G. J., 31 Chatterjee, A., 164 Freeman, G. T., 47
Airey, J., 190 Chung, W. K., 200, 221 Fuselier, Christopher S., 3,
Albertson, P., 50 Ciborra, C., 83 189
Alchian, A. A., 211, 212, Clark, K. B., 114
216 Clark, R., 201, 202
Gabarro, 1., 204
Allen, Thomas J., 5, 275, Coase, R. H., 207
Galbraith, J. R., 168
277,279,281,282, Cohen, H., 292
Geoffrion, A. M., 178
286 Cohen, Morris A., 3, 153,
Gerwin, D., 161
Andersen and Co., 68 164, 178
Gold,242
Anderson, O. E., Jr., 128, Coiffet, P., 16
Goldhart, J. P., 161
136 Cooper, A. c., 136, 149,
Graves, G. W., 178
Andrews, F. M., 279, 284 282,295
Grayson, C. J., 199
Anthony, R. N., 240 Cooper, W. W., 245
Argyris, c., 204 Copeland, T., 235
Asada, H., 24 Crawford, R. G., 216, 217 Hall, R. W., 166
Athos, A. G., 199,222,224 Hatvany, J., I, 45
Ayres, R. U., 51, 83 Davis, H.S., 242 Hayes, R. H., 162, 163,
Dearden, J., 240 167, 168, 221
Banker, Rajiv D., 4, 239, Deming, W. E., 68 Henderson, B., 234, 235
245, 246, 252 Demsetz, H., 211, 212, 216 Heron, S. D., 139, 149
Barney, Jay B., 4, 219, Denison, E. F., 200, 221 Heskett, J. L., 166
229, 231, 234 Dierdonck, R. Y., 165 Hogan, W. T., 142
Blackman, A. W., Jr., 140 Dilley, D. R., 142 Holusha, J., 222
Bolles, R. C., 31, 32 Douglas, R. W., 142 Hopkins, D. S., 157
Bornholtz, E. F., 70 Draper Lab Staff, 39, 40 Hunt, H. A., 18,84
Bourne, D. A., 48 Dupont-Gatelmand, c., 39, Hunt, T. L., 18, 84
Brown, J. W., 296 68,69 Hunt, Y. D., 22
Buijs,J., 157
Bureau of the Census, 107 Eikonix Corp., 17,50 Jablonowski, J., 39
Bylinski, G., 43, 44, 51 Engelberger, J. F., 20, 46, Janotik, A. M., 68, 69
50, 51, 82 Jellinek, M., 161
Cable,1. R., 216, 217 Engler, G. N., 140 Jenkins, R. Y., 136
Carnegie-Mellon University, Johnson, J. S., 68, 69
83 Farrell, M. J., 245 Jorgenson, D. W., 221
329
330 AUTHOR INDEX
Kan, A. H., 67 Nevins, J. L., 47 Schonberger, R. J., 68

Kanade, T., 24, 30 Nishimizu, M., 221 Seo, K. K., 222
Kanesaka, S., 47 Nitzan, D., 25, 27 Shephard, R. W., 245
Katz, Ralph, 5, 275, 277, Skinner, Charles S., 5, 301
279, 280, 282, 286 Okun, A. M., 204, 205 Skinner, W., 156, 159
Kearney & Trecker Corp., Oosterlinck, A., 31 Smith,286
38, 39, 51 Ouchi, W. G., 199,200, Smith, D. N., 85
Keller, S., 292 221,222,229,230, Sommers, W. P., 156
Kim, Linsu, 3, 113 231,233 Souder, W. E., 298
Klahorst, H. T., 39, 40, 51 Spence, A. M., 165
Klahorst, T., 41 Panzer, J. c., 160 Stauffer, R. N., 47
Klein, B. H., 148 Paris, M., 162 Stecke, K. E., 164
Klotz, T., 47, 51, 52 Parker, R. G., 67 Streeter, D., 292
Pascale, R. T., 199,222, Sumanth, 240
224
Lawler, E. E., 223
Paul, R. P., 16 Tarmann, M. B., 142
Lee, D., 99
Pelz, D. C., 279, 284 Tarondeau, J. C., 161
Lee, Hau L., 3, 153, 178
Perry, G., 47, 49, 51 Taylor, E. S., 140
Leibenstein, H., 203, 204,
Peters, T. J., 224, 285 Tech Tran Corp., 13, 19,22
210,211,212,218,
Porter, M. E., 156, 158, Thomas, R. R., 204
223
233, 234, 235 Thompson, H., 162
Levinson, H., 204, 208
Price, R. L., 221 Tomer, John F., 4, 199,
Lien, T. K., 47
Pugh, A., 28, 30 203, 208, 212, 222,
Likert, R., 212
223, 224, 229, 230
Little, A. D., 134, 135, 140
Rardin, R. L., 67 Tushman, M., 279
Luh, J. Y. S., 16
Rathmill, K., 1,45
Lund, R., 128
Rhode, J. C., 223 Utterback, James M., 3,
Riggs, J. L., 222 1l3, 114, 119, 120,
McBridge, D. L., 142 Rinnoy, 67 145, 278, 296
McNally, M., 220 Roberts, E. B., 275, 289,
Macneil, I.R., 219, 222 290,291,292 Van Boven, G. J. R., 31
Maindiratta, A., 245 Robotics in the UK, 83 Varian, H., 245
Maitai, S., 215 Romano, P., 83 Vietorisz, T., 76
Majumdar, B. A., 135 Romanosky, Jack, 4, 5, 259 Vliestra, J., 1
Manufacturing Engineering Rose, S., 142 Vollmann, T.E., 166
Magazine, 51 Rosso1, L., 27 von Hippe1, E. A., 149, 296
Marquis, D. G., 281
Maslow, A. H., 206, 221 Sadowski, R. P., 68, 69 Walther, J., 47
Maxwell, W., 164 Sanderson, A. C., 15,26, Warnecke, H. J., 47
Mayer, J. E., 99 47,49,51 Waterman, R. H., 224, 285
Mig1iarese, P., 83 Sanderson, R. J., 31 Wayne, K., 138
Miller, J. G., 165, 166 Schein, E. H., 219 Weiner, Steven A., 57, 67
Miller, L., 164 Schendel, D., 136, 149, Weiss, L. E., 15
Miller, Steven M., 2, 9, 42, 282,295 Weston, J. R., 235
50,51,73,83 Scherer, F. M., 164 Wheelwright, S. C., 162,
Miller, Steven S., 18, 20 Sch1aifer, R. D., 139, 149 163
Monden, Y., 68 Schmenner, R. W., 159, Whitney, D. E., 47, 83
Morey, R., 246 164, 167, 168 Wielinga, R. F., 1
Mueller, D. c., 224 Schoemaker, P. J. H., 168 Wiener, Steven A., 2
Murray, A. E., 145 Schon, D. D., 282, 297 Wilkins, A., 230
AUTHOR INDEX 331
Williamson, O. E., 158, Wilson, R. c., 85 Yoshikawa, H., 1,45

203, 207, 218, 230, Wright, P. K., 47, 49 Young, c., 190
231
Willig, R. 0., 160 Yale, J. P., 140 Zimmerman, N. J., 31

The Management of Productivity and Technology in Manufacturing PDF

Uploaded by

Copyright:

Available Formats

You might also like

The Management of Productivity and Technology in Manufacturing PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Management of Productivity and Technology in Manufacturing PDF

Uploaded by

Copyright:

Available Formats

The Management of

Plenum Press • New York and London

Main entry under title:

The Management of productivity and technology in manufacturing.

Includes bibliographical references and index.

ISBN-13: 978-1-4612-9516-7 e-ISBN-13: 978-1-4613-2507-9

Commissioned papers presented at the Wharton Conference on Productivity,

©1985 Plenum Press, New York

A Division of Plenum Publishing Corporation

All righ ts reserved

No part of this book may be reproduced, stored in a retrieval system, or transmitted

Thomas Allen • Sloan School of Management, Massachusetts Institute of

Jay B. Barney. Graduate School of Management, University of California,

Morris A. Cohen • The Wharton School, University of Pennsylvania,

Christopher Fuselier. General Electric Company, Industrial Automation

Ralph Katz • Graduate School of Business, Northeastern University,

Linsu Kim. Korea University, Seoul, Republic of Korea

Hau L. Lee. Department of Industrial Engineering and Engineering

Steven Miller • Graduate School of Industrial Administration, Carnegie-

Jack Romanosky. Arthur Andersen & Company, Philadelphia,

Charles S. Skinner • Strategic Technology Inc., Cleveland, Ohio

John F. Tomer • Department of Economics and Finance, Manhattan

James M. Utterback. Massachusetts Institute of Technology, Cambridge,

Steven A. Weiner. Ford Motor Company, Detroit, Michigan

This volume is concerned with the nature of new manufacturing

acknowledge the financial support of the Continental Group, Texaco

I. The Nature of New Manufacturing Technology

4.3. Developments in Flexible Manufacturing Systems ........ 46

II. Innovation and Manufacturing Strategy

2. A model of Innovative Challenges and Defensive Responses 121

III. Productivity Management

4. Concluding Remarks ........................................................... 253

IV. Technology Management and Organization

4.5. Organizational Slack and Sponsorship.......................... 296

Subject Index 317

Author Index 329

The chapters in this volume are concerned with four related

Little need be said to demonstrate the importance of these issues

1. THE NATURE OF NEW MANUFACTURING

Part I deals with the nature of the new manufacturing technolo-

innovations. The first chapter by Steven Miller discusses the nature of

2. INNOVATION AND MANUFACTURING STRATEGY

The traditional evolution of product and process interactions is

chapter by Utterback and Kim extends this discussion to include the

Following Ben Franklin's adage, "There is nothing so galling as

chapter by John Tomer, who concentrates on the incentives for

future, but there exist significant differences on the best means of

4. TECHNOLOGY MANAGEMENT AND ORGANIZATION

The intent of the conference where the above chapters were

Vlietstra, J., and Wielinga, R. F. (eds.), Computer-Aided Design. North-

INDUSTRIAL ROBOTICS AND

This chapter reviews some of the recent developments in

1. Extensions of the capabilities of commercially available robo-

Topics covered that relate to extensions of commercially avail-

Steven M. Miller. Graduate School of Industrial Administration, Carnegie-Mellon

machining in one plant, (2) the use of FMS in manufacturing

2. BACKGROUND ON ROBOTIC MANIPULATORS

2.1. The Robotic Manipulator