Professional Documents
Culture Documents
Effective Inventory and Service Management Through Product and Process
Effective Inventory and Service Management Through Product and Process
Effective Inventory and Service Management Through Product and Process
apid technology changes and increased globalization are two common characterizations of the environment faced by manufacturing enterprises of high
technology products. The immediate consequence of
such an environment is increased product proliferation.
Product proliferation creates a major operational challenge to managers of a manufacturing enterprise. It is
difficult to forecast demands accurately, leading to high
inventory investment and poor customer service.
Product proliferation is often unavoidable in a global
market. Different markets may have different requirements for the product, due to differences in taste, language, geographical environment, or government
regulations. For example, computer products for various countries may diEer in the power supply module to
accommodate local voltage, frequency, and plug conventions. Keyboards and manuals must match local
language. Telecommunication products may also be
differentiated by the communication protocols supported. In some cases, the need for localized versions
of a product results from government-imposed local
content requirements.
Expansion of the customer base may also lead to product proliferation. Different product versions are often developed for different market segments (e.g., education,
personal, business, or government users may have different needs of a product). Finally, rapid technology
changes mean that a company may be selling multiple
versions of the same product at the same time.
To deal with the operational problems of product proliferation, companies have invested in information technology, decision support systems, and transportation
modes. These investments aim at improving the
efficiency of the order fulfillment cycle. Another approach is to redesign the product or the process through
which the product is manufactured and distributed, to
gain control of inventory and customer service for the
product. While this approach is appealing, it often has
not been implemented (see Lee and Billington 1992).
Several obstacles exist (see Lee 1993 for more details),
which are outlined next.
First, design engineers have to take a broader perspective than product functionality, performance, and manufacturability (see Whitney 1988, and Dean and Susman
1989). It also requires enlarging the definition of costs in
the evaluation of alternative designs, which are often
narrowly defined as the material costs of the product and
direct labor for assembly.
Second, redesign of products for inventory and service
improvements often require close collaboration among
multiple functions, such as distribution, sales and marketing, finance, customer support, engineering, R&D,
and manufacturing, within a corporation. The organizational barriers between these functions can be very high.
Third, design changes may require some investment,
such as investing in additional manufacturing capabilities
at a distribution center, higher component costs, and additional vendor management costs.
Hence, management could be reluctant to proceed
with new designs that could result in inventory and service improvements, unless there is a concrete estimate of
the corresponding benefits. Logistic costs, such as
freight, customs and duties can easily be quantified. The
benefits in terms of lower inventory, faster response
times to customers, increased availability levels and increased flexibility, are much more difficult to quantify. It
Subject ckussificutio~zs: Inventoryiproduction, applications: product/process design; multi-echelon; supply chain management.
Area of review: MANUFACTURING,
OPER.~TIONS
AND SCHEDULING
(SPECIALISSUEON NEWDIRECTIONS
I N OPERATIONS
MANAGEMENT)
Operations Research
Vol. 44, No. 1, Janua~y-February 1996
0030-364X/9614401-0151$01.25
8 1996 INFORMS
1. LITERATURE REVIEW
DESIGN
In this section, we present two inventory models that can
be used to support product/process design. These two
models were all motivated by real application cases.
Both were used to support analysis of product/process
design to delay product differentiation. One deals with an
environment where an intermediate stage of the product
is built to stock, from which this intermediate product is
then customized into different final products on demand.
We term this a build-to-order model, because final products are built on demand. The other deals with an environment where immediate delivery of finished products
is critical, so that finished products are built to stock. We
term this a build-to-stock model. Both models assume
stationary demands and costs.
2.1. A Build-to-Order Inventory Model
Generic Production
Process
Different
Product
Versions
for
Customer
Intermediate lnventory
Stockpile (Generic Product)
+gr(t)
F(xlt) dx.
and
Prob{Y a R) = Prob{W
aR
Consider the scenario faced by a disc-drive manufacturer. Disc-drive orders come from different computer
manufacturers (OEMs), each of which orders a unique
set of products. Disc-drive manufacturing has a long lead
time, because many time-consuming tests are involved.
Hence, it is necessary to have in-process inventory to
shorten the response time to customer orders. High forecast errors resulted from high demand uncertainty implies that high levels of in-process inventory are needed
to provide a high level of reliability for meeting the target
response time to orders.
The manufacturing process of disc drive can be divided into two parts. All disc drives for any OEMs have
to first go through a generic part of the process. In
the second part, the disc drive is then customized to the
specifications of the different OEMs. Intermediate discdrive inventory is held at the end of the first part of the
process. The first part of the manufacturing process,
however, is relatively short. The second part of the process begins with some time-consuming tests that require
customized printed circuit boards.
Since the second part of the manufacturing process is
relatively long, a high level of intermediate inventory
stockpile ;s required to support high reliability of order
lead time targets. Based on the model analysis shown in
the last section, it would be ideal to design the process so
that the point of product differentiation is postponed
without increasing the value of inventory at that later
point. This calls for reconfiguring the process so that
process steps in the second part with little value added
can be made generic and so can be performed before the
differentiation point. It turns out that the testing steps
just described can be carried out using a "coupon"
board (can be viewed as a generic board) without the use
of the customized board on a drive. After the series of
tests are completed, the coupon board is then removed
and the actual printed circuit board is then inserted at
PCB
Insertion
end-product
Work-In-Process
lnventory Stockpile
Coupon Board
Insertion
PCB
Insertion
end-product
Series of Tests
Work-In-Process
lnventory Stockpile
Point of Product
Differentiation
Generic Production
Process
Customization
Process
as before: How to characterize the operational performance of such a system as a function of t , T, the required fill rate, and the demand distribution of the endproducts. Based on that, we can then explore the costs
and benefits of delaying the point of product differentiation t .
At the beginning of each time unit, the inventory status
of all end-products is reviewed, based upon which two
actions will be initiated. First, an allocation decision is
made regarding how items that have just completed the
first t time units of processing should be allotted to be
customized for different end-products. Second, the
amount of new items to begin production is determined.
Readers familiar with the distribution literature would
recognize that this manufacturing system resembles the
single warehouse, multiretailer system studied by Eppen
and Schrage (1981), Federgruen and Zipkin (1984), and
Schwarz (1989). Hence, we will only state the key operating characteristics without detailed proofs here. We
then present the results of our analysis based on these
characteristics.
Eppen and Schrage have shown that, using linear inventory holding and backorder costs and under fairly
mild assumptions, the optimal inventory policy is to operate each end-product stockpile as an order-up-to S system. Let S, be the order-up-to level for end-product i.
Hence, at the beginning of each time unit, the number of
new items to begin production is the total demand of the
previous time unit. Furthermore, an equal fractile allocation rule is used to allocate the inventory that has just
completed t time units of production for the customization to different end-products. This rule stipulates that,
after allocation, the inventory position for each endproduct should be the sum of the mean demand for the
end-product over T - t time units, and a common
safety-stock factor multiplied by the standard deviation
of demand for the end-product over the T - t time units.
Given such results, the steady-state end-of-period inventory level, I,, has a mean and variance given by (see
Eppen and Schrage, or Schwarz, for details):
xj$
8/
customers
Point of Product
Differentiation
-J
4-1.
In addition to inventory, there are many other factors
that should be considered in a comprehensive evaluation
of the localization alternatives. First, a localized printer
contains localization materials, and so it has a higher
value than an unlocalized printer. Hence, the capital tied
up in pipeline inventory (inventory in transit) is also
lower when localization occurs at the end of the chain,
i.e., DC localization. Second, an unlocalized printer is
much less bulky than a localized one, as the localization
materials and many of the final packaging materials that
are needed for the customers do not have to be bundled
with the printer. One can thus ship the unlocalized printers in bulk pallets, and cut the cost of transportation
significantly. Third, increasing "local content" and local
"manufa~turing~~
presence can make a product more
marketable, which supports doing localization at the nonU.S. DCs (see Cohen and Lee 1989). There is also a need
to develop a local supply base of the localization materials for the DCs. Finally, since DC localization requires
the DCs to perform some operations that are traditionally
viewed as manufacturing activities, there may be cultural
and organizational barriers to overcome. After considering both the quantifiable and the nonquantifiable factors
mentioned before, the manufacturer has redesigned the
product, and is designing all future products to support
the DC-localization strategy.
2.6. Another Application of the Build-to-Stock Model
Factory-LocalizationStrategy
mfg
DC
customers
Point of Product
Differentiation
DC-Localization Strategy
1.
2.
3.
4.
No Common
Key Parts
Common Head
Driver Board
Head
Driver Board,
Common Print
Mech Interface
, color printer
*
*
Common PCB
Assembly
PCB Assembly
color printer
mono printer
Common PCB
~
mono printer
Common
~
~i~ ~
~~ ~ l
color printer
~ mono
~
printer
b ~
Final Assembly
3. CONCLUSIONS
The model described in subsection 2.4 is useful for estimating the cost item of item 4.
l ~
~ b