Manufacturers in many industries are
facing the supply chain challenges of product
proliferation [1]. For global firms, specific
local requirements such as language,
conventions, and government regulations
mean that any single product must have
multiple product derivatives. In the U.S.,
market segmentation by business and
consumer channels, price, and feature set
further increase product variety. For example,
consumer electronics and PCs are often
customized for each retail channel allowing
Wal-Mart to sell a slightly different product
than Best Buy or Office Depot. Satisfying the
customization needs of each channel creates
many supply chain complexities as
manufacturing and distribution struggle to
manage a wide range of product derivatives
[2]. Even forecasting the volume of multiple
niche products is ever more difficult. Many
supply chains rely on large inventory holdings
to reduce the risk of poor product availability.
However, this is costly and unsustainable in
highly competitive markets.
To further complicate the situation,
technology advances have shortened the life
cycles for many products, especially in
electronics and computers. The short life
cycles drastically increase the penalty of
holding obsolete finished goods inventory.
For example, at Hewlett-Packard, the average
life cycle for a DeskJet printer product is
approximately 18 months, with some
derivatives lasting only a few months. In PCs,
six months is more typical, with some
products lasting only a few weeks in the
channel! The annual cost of holding
inventory of printers or PCs may approach
50% of the product cost since products lose
value every day and old products must be
deeply discounted or sold through alternative
channels. Moreover, in computer and
peripheral markets, manufacturers face
constant price competition and narrowing
margins, requiring both low inventories and
high service levels to ensure profitability on
product development investments.
Manufacturers in such industries have
developed many supply chain strategies to
address the problems that accompany
product proliferation. Delayed product
differentiation or postponement is one
approach that has proven to reduce inventory
needs while ensuring high product
availability [3]. In this paper, a model is
developed to explore the value of
postponement under different operating
conditions. This model was developed to
better understand when design investments
that facilitate postponement are most
Postponement Strategies for
Channel Derivatives
M. Eric Johnson
Dartmouth College
Emily Anderson
Agilent Technologies
Volume 11, Number 1 2000 Page 19
The value of postponing product differentiation until final distribution for
manufacturers who market a family of product derivatives through multiple channels
is examined. A model is developed of a supply chain that distributes many
short-lived products through different channels. Using the model, we find the
postponement is particularly valuable for managing short-life products.
Postponement increases distribution service levels while reducing costs and order
fulfillment risk. Postponement is particularly valuable when there are many
derivative products and forecast error is high. Trade-off curves are presented, that
allow managers to evaluate the benefits of investing in postponement strategies.
The annual cost of
holding inventory of
printers or PCs may
approach 50% of the
product cost since
products lose value
every day and old
products must be
deeply discounted or
sold through
alternative channels.
J/A article.qxd 9/20/01 9:35 AM Page 19
beneficial. Specifically, using this model
we examine design strategies
where manufacturers invest in product
platforms [4] that can be easily customized
into many different product derivatives. The
development cost of such platforms and the
added material cost to each product can be
significant [5]. Thus, we seek to understand
when these investments are warranted.
Because manufacturers face tradeoff
decisions around product postponement and
flexible factories, the sensitivity results from
the model show how postponement decisions
can be effected by different operating
conditions - and under which conditions
postponement provides the most benefit. The
operating conditions focused on are inventory
policy, forecast uncertainty, product variety,
product mix, and postponement premium.
Additionally, the impact of postponement on
order fulfillment risk is evaluated. The model
and the results provide a management tool for
predicting the impact of postponement on
future product platform introductions.
Literature Review
The concept of postponing product
differentiation beyond manufacturing has
been discussed for over 50 years [6], [7].
However, it was only about ten years ago that
logistics researchers began to define and
study the concept [8]. In the past five years,
the demands of managing global product
offerings have pushed managers in many
industries to seriously consider postponement
as a supply chain strategy for mass
customization [9]. This has renewed
researcher interest in studying the benefits of
postponement.
In their landmark paper, Zinn and
Bowersox [10] defined and analyzed five
different types of postponement (labeling,
packaging, assembly, manufacturing, and
time). Using simulation models, they
examined conditions that favor the different
types of postponement. Hewlett-Packard
reported one of the early successful
applications of postponement in the
computer industry involved localizing
products for global markets [11]. HP
manufactured printers in the U.S. and
distributed finished products globally through
three distribution centers in Europe, the U.S.
and the Far East. Each country had their own
local requirements including the appropriate
power supply module, power cord
terminators, and manuals in the appropriate
language. Previously, localization was done
in the U.S. factory and finished products were
shipped to the three distribution centers
(DCs). However, the long transit times to the
DCs required them to hold high levels of
safety stock. HP began to investigate the
benefits of a product and process redesign
where a generic printer would be produced at
the factory and shipped to the DCs for final
customization with the power supply and the
manual. The printer itself had to be
redesigned so that the power supply model
could be added easily at the DC, which
required some additional investment at the
DC to give them this capability. The results
from the DC localization at HP were positive
- inventory costs were reduced while a
customer service measure like fill rate
improved. The value of the pipeline (or in
transit) inventory was lower because it was in
a generic form and the unlocalized printers
were less bulky and therefore less costly to
ship. HP also observed that increasing the
local content and local manufacturing
presence made the product more marketable.
This success story strengthened the industrial
interest in postponement, motivating further
research including the research described in
this paper.
Using an analytical model, Lee [12]
examined how product and process redesign
for delayed product differentiation
(postponement) could be used to improve
inventory and service management. Lee
examined disc drive manufacturing, which
typically required long lead-times due to
lengthy testing. He developed two inventory
models that could be used to support product
and process redesign decisions. Lee found
that value of delayed differentiation was
greatest when the process was designed so
that customization took place after long, non-
value-added steps were performed. For
example, Lee described how disc drive
manufacturers could use a generic coupon
board during testing and then insert a
customized printed circuit board during final
assembly – postponing the final configuration
of the disk drive.
Lee and Tang [13] pointed out that
before redesign for localization or
Page 20 The International Journal of Logistics Management
In the past five years,
the demands of
managing global
product offerings have
pushed managers in
many industries to
seriously consider
postponement as a
supply chain strategy
for mass
customization.
J/A article.qxd 9/20/01 9:35 AM Page 20
customization is initiated, the economics
must be analyzed because some fixed and
variable costs associated with a
product/process redesign will change. They
analyzed three basic approaches to delayed
product differentiation and discussed the
conditions that resulted in the greatest benefit.
The first was standardization of components,
whereby a part was designed to be common
to all products. Lee’s multi-stage model
captured the additional material and
processing costs that result from
standardization and the costs of holding
buffer inventory at intermediate stages in the
product process. The model could also be
used to evaluate an optimal stage of the
process for part standardization. They were
able to show that standardization was
effective only when the investment and
processing costs required for standardization
were low. The second approach they
evaluated was modular design, where a part
was divided into two modules. The first
module was common to all products and the
assembly of the second module was deferred.
Lee and Tang found that with this approach it
paid to delay the product differentiation from
stage 1 to stage 2 when the lead-time of stage
2 was long, when the additional module was
easy to handle or when the modular design
was relatively inexpensive. The third
approach was process restructuring where the
product differentiation results from
postponing an operation downstream in the
supply chain. They found that this approach
was beneficial when the lead-time for the first
common stage was long and when stage 2
(postponed step) was a high value-added
activity. Lee and Tang [14] also investigated
process restructuring through operations
reversal whereby the manufacturing process
was reengineered and two consecutive stages
of the process were reversed. This provided
the greatest benefits when high value-added
activities were deferred.
Recently, Pagh and Cooper [15]
developed a framework for describing different
postponement and speculation strategies,
while Mason-Jones and Towill [16] considered
the role of postponement approaches in
decoupling information and material in the
supply chain. For an extensive review of the
literature on postponement, see “The Benefits
of Design for Postponement” [17].
Compared with the earlier research on
postponement, this paper concentrates on
cases where demand is non-stationary. Most
of the previous work concentrated on
products, whose demand was uncertain, yet
the average demand values were constant.
We examine products that experience short
life cycles where demand typically surges at
product introduction; remains steady for a
few months, and then decline at the end of
life. Such demand patterns are typical for
products in the computer and electronics
industry along with other fashion industries
like apparel and toys. We also examine
additional elements representative of many
supply chains including multiple channels,
each facing its own demand distribution and
complex cost structures including
postponement premiums related to material
and design costs.
Model Description
To examine the issues around platform
design for channel derivatives, we developed
a model to evaluate the benefits of
postponement under a variety of different
operating conditions. The model represents a
two-level supply chain where a single
manufacturing plant supports multiple
demand channels, each facing non-stationary
demand from end customers. These different
channels could represent similar products
sold through multiple consumer channels
such as Wal-Mart, Office Depot, and Best Buy
or alternative commercial channels such as
Ingram Micro. They could also represent
direct e-channels like HP.com or
pcorder.com. For each case, we compared
the inventory and service performance of a
traditional supply chain (see Figure 1) where
multiple products are produced at the factory
to a supply chain where a generic product is
produced at the factory and then later
differentiated in the distribution channel (see
Figure 2).
We used this model to analyze the
inventory costs that distribution would incur if
those products or channels were
manufactured individually (non-
postponement), or differentiated from a single
platform downstream in the supply chain
(postponement). Because manufacturers face
many trade-off decisions around product
postponement, the objective was to
Volume 11, Number 1 2000 Page 21
J/A article.qxd 9/20/01 9:35 AM Page 21
determine how beneficial postponement
would be under a variety of operating
conditions. The conditions we focused on
were:
• Product variety, or the number of
derivatives from a single platform.
• Product mix - symmetry or asymmetry of
demand between derivative products.
• Inventory policy decisions.
• Service level goals (fill rate).
• Forecast error.
• Postponement premium cost (material and
processing costs).
Model Input
Figure 3 shows a flowchart of the
activities that occur each period in our
simulation model. The simulation was
initiated with the product forecasts and all
other system parameters. Product demand for
each channel was generated as a normal
distribution using the forecast as the mean
Page 22 The International Journal of Logistics Management
Figure 1
Traditional Supply Chain Where Manufacturing Produces Final Products
4
3
2
1
4
3
2
1
4
3
2
1
4
3
2
1
Channel 1
Manufacturer
Channel 2
Customers
Customers
Figure 2
A Supply Chain Where Manufacturing Produces a Generic Product that is Configured by
Distribution to a Channel Customized Offering
4
3
2
1
4
3
2
1
4
3
2
1
Channel 1
Manufacturer
Channel 2
Product Platform
at DC
Multiple SKUs
Customers
Customers
J/A article.qxd 9/20/01 9:35 AM Page 22
and the forecast error as the standard
deviation. Inventory planning for the non-
postponement cases was based on the
forecast for each product derivative
(individually). For postponement, an
aggregate platform forecast was used for
planning. After determining the beginning
inventory, prior period orders were received
and then the demand was filled from the
inventory stockpile. If the demand exceeded
the inventory stockpile, then demand was
either lost or marked as backordered,
depending on the backorder parameter
(percentage of customers willing to wait). An
order quantity was calculated based upon the
ending inventory, a foreword look to the
forecast, lead-time from the factory, and a
desired safety stock policy. Replenishment
orders were then recorded as pipeline
inventory. Replenishment orders were not
taken out of the pipeline until they were
received at the distribution center. After the
demand was filled in any period, the
remaining inventory was assessed a holding
cost per unit.
The user input variables to the model
were:
• Number of derivative products on a
platform.
• Product forecast and forecast error.
• Lead-time from the factory to the
distribution center.
• Desired safety stock policy.
• Platform cost.
• Postponement premium.
• Holding costs.
• Cost of a lost sale (including possible lost
sales on future consumables related to the
product).
The postponement premium was
specified as a percentage of product cost
incorporating both the costs of common
platform development and postponement
production and materials.
Model Output
The key performance metrics captured
by the model was supply chain costs and
service level [18]. The costs included
Volume 11, Number 1 2000 Page 23
Figure 3
Steps in the Simulation for Each Period
Input Date
(forecasts, parameters, costs)
Update
Beginning Inventory
Update
Ending Inventory
Update
Pipeline Inventory
Calculate
Order Quantity
Calculate
Holding Cost ($)
Calculate
Lost Sales ($)
Inventory
Shortage?
Record Lost Sales
Record Backorders
Calculate Quantity Received
Match Qty Ordered with
Period Due
Determine:
Order Received?
Match order with
period due
Generate Demand
YES
YES
NO
NO
The key performance
metrics captured by the
model was supply chain
costs and service level.
J/A article.qxd 9/20/01 9:35 AM Page 23