Professional Documents
Culture Documents
Selecting Winning New Product Projects: Using The Newprod System
Selecting Winning New Product Projects: Using The Newprod System
Selecting Winning New Product Projects: Using The Newprod System
1985;2:34-44
oc>oo
Robert G. Cooper
No one screening method available today meets all relatively simple consumer products, such as packaged
the requirements for an ideal screening tool as outlined goods. Such techniques assume that the critical criteria
above. Instead, a variety of methods have been devel- for the GO/KILL decision solely involve the mar-
oped, and each incorporates some of the desired fea- ketplace; technological and production issues are either
tures. The four main approaches to initial screening obvious or simple to solve. Given a market-based
include [2,5,17]: screening decision, it makes sense to use a variety of
market research techniques, ranging from consumer
1. Benefit measurement models.
panels and focus groups to perceptual and preference
2. Economic models.
mapping, to screen product ideas.
3. Portfolio selection models.
Of all the dozens of screening models proposed,
4. Market research approaches.
benefit measurement approaches are generally recom-
Benefit measurement models require a well-in- mended for new product idea screening. Because only a
formed respondent or group to provide subjective in- tentative commitment is required, and since available
puts regarding characteristics of the project under con- information on the project is limited, benefit models
sideration [3]. Such models are systemmatic become the logical screening tool [23]. For example,
procedures for soliciting and integrating benefit data. the Conference Board reports that about half the firms
They typically avoid conventional economic inputs, they studied have set forth writen guidelines or rules for
but rely more on subjective assessment of proxies for project selection [ 161. Usually these are in the form of
new product success and payoffs, and fit with corporate checklists or scoring models. Souder investigated 26
objectives. Included in this category are checklists, and project selection models and found that managers rated
their extension, scoring models. In the latter, ratings of scoring models best in terms of ease of use and cost to
a project’s attributes are sought and combined in a implement [23]. Souder concludes that scoring models
weighted fashion to yield a numerical project score. are “highly suitable for preliminary screening deci-
Economic models treat the idea screening decision sions where only gross distinctions are required among
much like a conventional investment decision. Com- projects. ”
putational approaches, such as payback period, break-
even analysis, return-on-investment, and discounted
cash flow methods, are used. To accommodate the un-
certainty of data, techniques of probability, including Scoring Models: Their Shortcomings
Monte Carlo simulation, risk analysis, and decision
tree analysis, have been proposed. But at the idea
screening stage, economic models suffer because they
require considerable financial data as inputs when In spite of their popularity, scoring models are plagued
often relatively little is known about the project. Thus, by difficulties [20]. Such models rely on the subjective
such models are usually considered more relevant for ratings of managers, and hence data input may not be
“known” projects (line extensions, product modifica- very reliable. However, at the screening stage, man-
tions, etc.), or at later stages of the new product agement opinion is often the only data available. More-
process. over, ratings from several evaluators together with con-
Portfolio selection models view the screening deci- fidence scores can be combined to yield a composite
sion as part of the total resource allocation problem. and more reliable value for each input variable. The
Such techniques largely involve operations research premise here is that the “average” decision maker is
constrained optimization methods, such as linear, in- near optimal; unfortunately, none of us is ‘ ‘average. ’ ’
teger, and dynamic programming [ 131. The objective is Other criticisms tend to be of a technical nature.
to develop a portfolio of new and existing projects in Often, scoring models are seen as oversimplifications,
order to maximize an objective function, yet subject it since they attempt to reduce a complicated decision
to a set of resource constraints. Because these mathe- situation to a product score [20]. A major deficiency is
matical models require substantial data inputs, includ- the arbitrariness of items or checklist questions used,
ing financial data on all projects, timing information, and that importance weightings assigned to each criteri-
resource needs, and availabilities, they are rarely used on are also arbitrarily determined. The selection of
PI. these questions and weights are no doubt based on the
Market research approaches are usually limited to judgment and past experience of the model developer.
SELECTING WINNING NEW PRODUCT J PROD INNOV MANAG 37
1985;2:34-44
However, his or her experience may be limited to a The difference is that NewProd was derived from a
handful of projects, while his/her ability to judge how large number of past new product successes and fail-
important each item is and to translate these into numer- ures. In short, the questions or items used and the
ical weights, may be limited. Freimar and Simon pro- weights attached to these items are based on a statistical
pose the use of linear discriminant analysis of a large analysis of almost 200 projects from 100 companies.
number of past successes and failures in order to identi- And the model has been validated, yielding a predictive
fy the weights to attach to each screening variable [20]. ability-ability to predict accurately a product success
Another weakness is the fact that many of the vari- or failure-of approximately 84%.
ables or items are not independent of each other. For
example, if one of the screening items is “com-
patibility with distribution channels,” then it certainly The Development of NewProd
is not independent from the “compatibility with cur-
rent products” measure. (Both items are taken from Here’s how the NewProd model was derived. A total of
O’Meara’s model [ 181). The result is that certain items 195 industrial new product cases from 102 firms was
are double counted. Schocker, Gensch, and Simon note identified. Half of these were commercial successes,
that factor analysis of the many screening variables to a i.e., met or exceeded the acceptable financial return for
subset of independent factors or dimensions could be this type of investment. The other half had been
used to eliminate the interdependence of ratings [21]. launched, but were subsequently rated as commercial
failures.
For each project, managers were asked to rate the
project on each of 80 characteristics. Zero-to-ten rating
The Advantages
scales were used. Of the 80 characteristics, a total of 48
In spite of these criticisms, the scoring model is perhaps were judged to be potentially useful screening criteria,
the best idea screening tool available. Proponents argue since they would be known at the outset of the project.
that such a model has utility for a number of reasons: it The 48 characteristics described such features of the
helps make a highly judgmental decision somewhat project as:
more objective; it systematizes the review of projects; it
forces managers to subject each project to a consistent its marketplace; for example, market size,
and large set of review criteria; it focuses attention on growth, level of competition, etc.
the most relevant issues; it requires management to
state goals and objectives clearly; it is easy to under- the product advantage; for example, product
stand and use; and it is generally applicable.
uniqueness, quality, superiority, etc.
Regression
coefficient
(weight of
Key factors or dimensions (factor name) factor) F value Variables or items loading on factor
product superiority, quality, and uniqueness 1.744 68.7 product is superior to competing products
product has unique features for user
product is higher quality than competitors
product does unique task for user
product reduces customers’ costs
product is innovative-first of its kind
overall project/co. resource compatibility 1.138 30.0 a good “fit” between needs of project and company
resource base in terms of:
managerial skills
marketing research skills
salesforceidistribution resources
advertising/promo resources
financial resources
engineering skills
R&D resources
production resources
market need, growth, and size 0.801 12.5 high need level by customers for product class
large market ($ volume)
fast growing market
economic advantage of product to end user 0.722 10.2 product reduces customers’ costs
product is priced lower than competing products
newness to the firm (negative) -0.354 2.9 project takes the firm into new areas for firm such as:
new product class to company
new salesforceidistribution
new types of users’ needs served
new customers to company
new competitors to company
new product technology to firm
new production process to firm
technological resource compatibility 0.342 2.5 a good “fit” between needs of project and company
resource base in terms of:
R&D resources & skills
engineering skills & resources
constant 0.328
advertising and promotion resources, financial re- cases, but commitment to the project was so strong
sources, R&D and engineering talents, and production that it was launched in spite of the negative prognosis;
capabilities. and the outcome was as predicted, i.e., a strong
Newness to the Firm is a negative factor for project failure .)
outcomes. Projects involving a new product class to the
company, new distribution and salesforce approaches,
new types of customer needs, new advertising and pro-
Developing Your Own Model
motion methods, a new clientele, new competitors,
new product technology, or a new production process The methodology that we employed to derive the New-
for the firm fare more poorly. Prod model can also be used by organizations in order
The final important project-company fit dimension to develop their own screening models. The basic steps
is Technological Resource Compatibility, a positive are:
factor. Technologically compatible projects are those
1. Develop a reasonable set of screening items or ques-
that fit well into the firm’s existing base in terms of
tions. The items we used in NewProd are probably a
R&D resources and skills, and engineering resources
good place to begin.
and skills. But note that production and technological
2. Identify a sample of past new product successes and
newness (also uncovered in the factor analysis)-pro-
failures within the corporation.
jects involving new production processes and technolo-
3 Request one or more evaluators to rate each of these
gies to the firm-has no impact on product outcomes.
past projects on each of the criteria developed in
item 1.
4. Using appropriate statistical techniques, e.g., factor
Other Criteria analysis, multiple regression, and/or linear discrim-
inant analysis, derive a subset of key underlying
A group of factors were identified that we later called
dimensions and a success equation.
‘ ‘project descriptors. ’ ’ Most of these, including Prod-
5. Validate the model. Use either a cross split half
uct Technical Complexity and Product Determinate-
method (described above) or better yet, some new
ness (how clear the product specification and tech-
test cases (as we have subsequently done).
nical solution are at the outset of the project), have no
6. Develop the computer software to handle the eval-
link to product outcomes. Only one entered the success
uator’s inputs. (Both batch and user friendly in-
equation, but in a fairly weak manner, namely Product
teractive programs for an IBM-PC have been writ-
Scope: custom products aimed at one or a few custom-
ten for NewProd.)
ers fare more poorly than do mass-market, broad-ap-
7. Establish a procedure within the firm to facilitate the
peal, more standardized new products.
use of the model, beginning with a teaching session
in a seminar format.
Mean
weighted Std.
Factor name Evaluator 1 Evaluator 2 Evaluator 3 evaluator Deviation
Table 3. Factor Scores and Impacts: Factor Table 4. Overall Project Ratings: Project Scores
Impact Table and Probabilities
played. These high variance items become the ority was the product rated poorly. And this one factor
topics of the meeting agenda. The meeting fo- proved to be a damaging one to the overall assessment
cuses on questions such as: why was there so of the project. Discussions ensued and the R&D project
much disagreement amongst us on item number manager confessed that it had been his goal to develop a
X or why did evaluator A rate item Y so high, and product that would simply equal a competitor’s prod-
evaluator B rate that item so low? uct. The outcome of the debriefing meeting was a rec-
Usually the reasons for the disagreement are ognition that product superiority was essential to the
readily identified. For example, in one evaluation product’s success. And two studies were immediately
of an industrial chemical, two of the evaluators commissioned to achieve this: an end-user interview
assumed a positioning strategy and niche target study to identify weaknesses in the competitive product
market that avoided a head-on confrontation with as perceived by potential customers; and in-house
major competitors; the other three evaluators creativity sessions that focused on ways that the pro-
rated the project assuming a nose-to-nose posi- posed new product could be significantly improved.
tioning strategy with competitors. The input rat- Thus the output of a well-conceived screening model is
ings and resulting prognoses were quite different not only a GO/KILL evaluation, but most importantly,
between these two groups of evaluators. But the an indication of what needs to be done next in the event
reasons for these differences became quickly ap- of a GO or tentative project.
parent, and a discussion on positioning strategy A final benefit of a systematic screening procedure is
ensued. The product was eventually launched, the payoffs from involving evaluators from different
targeted at the niche segment, and as predicted, functional areas within the firm. In one company, typ-
became a success. In other cases, discussion of ical new product projects had been screened and re-
the inputs that feature high variances among eval- viewed by an R&D committee. Not surprisingly, there
uators reveals that one or more evaluators has had had been many complaints from the operating divisions
first-hand knowledge that he or she then shares about the types of new products that eventually found
with the group. So the debriefing session be- their way out the lab and into the divisions. The intro-
comes a knowledge-sharing forum, with the duction of the NewProd system forced the involvement
group moving towards a common knowledge of people from the operating divisions in the screening
base. This sharing of experience is particularly decision, but in a relatively painless and efficient man-
fruitful when the group is multidisciplinary. ner. Not only did better screening decisions result from
these multidisciplinary inputs, but a thorny political
l The profile of the project on each of the eight key problem was solved.
factors helps to identify and prioritize the pros
and cons of the project. The relative impacts-
positive or negative-of these factors provide
key insights into needed action. Three typical Learning from Experience
questions on a negative evaluation, for example,
include the following:
Why did the project come out relatively negatively? New product success can never be guaranteed. Indeed
What were the “killer factors”? And most important, new product development is perhaps the riskiest en-
can we do anything to improve these negative factors? deavor of the modem corporation. But new product
development must continue if the corporation is to
prosper.
Project selection is pivotal to effective risk reduction
A Planning Tool
in new development. A scoring model, such as New-
The outcome of this impact analysis is very often a Prod, can be a valuable tool in screening new product
study or task to correct a highly negative feature of the proposals. Moreover the diagnostics of such an evalua-
project, or to confirm a critical positive feature. tion help guide the project by identifying key questions,
For example, in an evaluation of new building mate- issues, and tasks to be undertaken.
rial, the product fared moderately positive on most of We can always learn from our past triumphs and
the eight screening factors. Only on Product Superi- disasters. An empirically derived screening model,
44 JPRODINNOVMANAG R.G. COOPER
1985;2:34-44
based on these past experiences, is simply a convenient 11. Cooper, Robert G. The impact of new product strategies. Zndustrial
Marketing Management 121243-256 (1984).
way of translating these experiences into a management
12. Crawford, C. Merle. New product failure rates-facts and fallacies.
decision tool, which promises to yield better results Research Management 9-13 (Sept 1979).
than traditional approaches. 13. Gear, A. E., Lockett, A. C., and Pearson, A. W. Analysis of some
portfolio selection models for R&D. IEEE Transactions on Engineer-
ing Management EM-1866-76 (May 1971).
References 14. Glotskey, G. R. Research on a research department: an analysis of
1. Albala, America. Stage approach for the evaluation and selection of economic decisions on projects. IEEE Transactions on Engineering
R&D projects. IEEE Transactions on Engineering Management Management EM-7:166-172 (Dee 1960).
EM-22:153-162 (Nov 1975). 15. Hopkins, David S. New Product Winners and Losers. Conference
2. Baker, N. R. R&D project selection models: an assessment. IEEE Board Report #773, 1980.
Transactions on Engineering Management EM-2 1: 165% 17 1 (Nov 16. McGuire, E. Patrick. Evaluating New-Product Proposals. New York:
1974). Conference Board Report No. 604, 1973.
3. Baker, N. R. and Freeland, James. Recent advances in R&D benefit 17. Moore, John R. Jr. and Baker, Norman R. Computational analysis of
measurement and project selection methods. Management Science scoring models for R&D project selection. Management Science
21:11&1-1175 (1975). 16:B212-232 (Dee 1969).
4. Baker, N. R. and Pound, W. R&D project selection: where we stand. 18. O’Meara, J. T. Jr. Selecting profitable products. Harvard Business
IEEE Transactions on Engineering Management EM- 11: 124- 134 Review 83-89 (Jan-Feb 1961).
(1964).
19. Rubenstein, A. H. Research project selection (an industrial researcher’s
5. Booz, Allen and Hamilton. New Product Managementfor the 1980’s. view). Research Management 71225-233 (May 1964).
New York: Booz, Allen and Hamilton, 1982.
20. Simon, L. S. and Freimar, M. Analytical Marketing. New York: Har-
6. Calantone, Roger and Cooper, Robert G. New product scenarios: pros- court, Brace & World, Inc., 1970.
pects for success. Journal of Marketing 45 (Spring 1981).
21. Schocker, A. D., Gensch, D., and Simon, L. S. Toward the improve-
7. Charles, A. and Stedry, A. C. Chance constrained model for real-time ment of new product search and screening. In: 1969 Fall Conference
control in research and development management. Management Sci- Proceedings, Amer. Marketing Assoc., P. R. McDonald (ed.). 1969,
ence 12:B353-62 (April 1966). pp. 168-175.
8. Cooper, Robert G. Identify industrial new product success: project 22. Souder, W. E. A system for using R&D project evaluation methods.
NewProd. Industrial Marketing Management 8: 124- 135 (1979). Research Management 21129-37 (Sept 78).
9. Cooper, Robert G. The dimensions of industrial new product success 23. Souder, W. E. A scoring methodology for assessing the suitability of
and failure. Journal of Marketing 43 (Summer 1979). management science models. Chemical Engineering 123- 126 (Ott 28,
10. Cooper, Robert G. An empirically derived new product project 1963).
selection model. IEEE Transactions on Engineering Management
EM-28:54-61 (Aug 1981).