Description
Supply chain management is likely to play an important role in the digital economy. In this paper, we first describe major issues in traditional supply chain management.
Models for Supply Chains in E-Business
Jayashankar M. Swaminathan • Sridhar R. Tayur
Kenan-Flagler Business School, University of North Carolina, Chapel Hill, North Carolina 27599
Graduate School of Industrial Administration, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213
S
upply chain management is likely to play an important role in the digital economy. In
this paper, we ?rst describe major issues in traditional supply chain management. Next,
we focus our attention on the supply chain issues of visibility, supplier relationships, dis-
tribution and pricing, customization, and real-time decision technologies that have risen to
importance with the prevalence of e-business. We present an overview of relevant analytical
research models that have been developed in these areas, discuss their contributions, and
conclude with a discussion on future modeling opportunities in this area.
(Supply Chain Management; Electronic Business; Collaboration; Information Sharing; Decision
Support; Supplier Relations; Procurement; Distribution; Customization; Literature Survey)
1. Introduction
It is estimated that e-commerce in the United States
will grow from $72 billion in 2002 to approximately
$217 billion by 2007, according to a recent Forrester
research report (Johnson et al. 2002). While the pre-
dicted numbers may not be exact, there is no doubt
that the rapid growth and adoption of the Internet
has already had a great impact on all aspects of
business, including customer acquisition, marketing,
human resource management, ?nance, information
systems (IS), and operations. Supply chain manage-
ment that has played an important role in traditional
businesses is likely to be crucial in the digital econ-
omy as well (Geoffrion and Krishnan 2001). Keenan
and Ante (2002) note that during the next ?ve years,
collaboration by supply chain partners over the Inter-
net can potentially save $223 billion with the reduc-
tion in transaction, production, and inventory costs.
This could be one of the largest bene?ts of Internet
technology and many ?rms have already begun to
realize these bene?ts. For example, Microsoft used
a Web collaboration tool to bring the Xbox video
game console to market two months ahead of sched-
ule (Keenan and Ante 2002); Cutler-Hammer has dou-
bled pro?ts and increased productivity by 35% for its
con?gured motor control centers and control panels
by coordinating with customers over the Internet
(Bylinski 2001); and Autoliv reduced the plant inven-
tory by 37% by coordinating orders online with sup-
pliers (Lundegaard 2001). The large potential impact
of the Internet on supply chain management makes
the study of supply chain models in e-business timely
and important.
Supply chain management is a vast topic. We
?rst provide a comprehensive de?nition of supply
chain management and then discuss opportunities
and changes created as a result of Internet usage.
1.1. Supply Chain Management
A supply chain is the set of entities involved in
the design of new products and services, procuring
raw materials, transforming them into semi?nished
and ?nished products, and delivering them to the
end customer. This de?nition, or a modi?ed ver-
sion of it, has been used by several researchers (see
Lee and Billington 1993, Swaminathan et al. 1998,
Swaminathan 2001a, Keskinocak and Tayur 2001).
Supply chain management is the ef?cient manage-
ment of the end-to-end process, which starts with
the design of the product or service and ends with
the time when it has been sold, consumed, and
?nally, discarded by the consumer. This complete pro-
cess includes product design, procurement, planning
0025-1909/03/4910/1387
1526-5501 electronic ISSN
Management Science © 2003 INFORMS
Vol. 49, No. 10, October 2003, pp. 1387–1406
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
and forecasting, production, distribution, ful?llment,
after-sales support, and end-of-life disposal. Supply
chain management issues can be classi?ed into
two broad categories: con?guration (design-oriented)
issues that relate to the basic infrastructure on
which the supply chain executes, and coordination
(execution-oriented) issues that relate to the actual
execution of the supply chain.
Con?guration-level issues include the following
topics.
Procurement and Supplier Decisions. How many
and what kinds of suppliers are necessary? Which
parts should be outsourced and which should be kept
in house? How can procurement practices be stream-
lined and standardized? How should long-term and
short-term contracts be used with suppliers?
Production Decisions. In a global production net-
work, where and how many manufacturing sites
should be operational? How much capacity should be
installed at each of these sites? What kinds of prod-
ucts and services are going to be supported through
the supply chain? How much variety should be pro-
vided to customers? What degree of commonality is
required across the product portfolio?
Distribution Decisions. What kind of distribu-
tion channels should a ?rm have? How many and
where should the distribution and retail outlets be
located? What kinds of transportation modes and
routes should be used? How should a ?rm exploit
risk-pooling opportunities?
Information Support Decisions. Should enterprise
resource planning software be standardized across
the functional units of a ?rm? Should the supply
chain work on standard protocols or on proprietary
standards?
Coordination level issues include the following
topics.
Material Flow Decisions. How much inventory of
different product types should be stored to realize the
expected service levels? Should inventory be carried
in ?nished form or semi?nished form? How often
should inventory be replenished? Should a ?rm make
inventory decisions or is it better to have the vendor
manage the inventory? Should suppliers be required
to deliver goods just in time?
Information Flow Decisions. In what form is
information shared between different entities in the
supply chain: paper, voice via telephone, electronic
data interchange (EDI)? How much collaboration
occurs among the supply chain partners during new
product development?
Cash Flow Decisions. When do suppliers get paid
for their deliveries? What prices should be charged
for products? What kinds of cost reduction efforts
are taken across the supply chain (or expected of
suppliers)? In a global ?rm, in which currency will a
supplier be paid?
It is clear that supply chain management spans sev-
eral functional and geographical areas, introducing
complexities both in terms of design and execution.
Some of the pertinent factors that complicate supply
chain management decisions include the presence of
multiple agents and their sometimes con?icting incen-
tives; uncertainty in demand, supply and produc-
tion distribution process; asymmetry of information
related to product design, inventory, costs, demand
and capacity across the supply chain between the
various parties in the supply chain; and lead time
between the different entities in the supply chain.
All these complexities lead to several types of inef-
?ciencies in the supply chain: poor utilization of
inventory assets (Lee and Billington 1992); distor-
tion of information such as the bullwhip effect (Lee
et al. 1997); high stock-outs and nonresponsive sup-
ply chains (Fisher et al. 1994); poor customer ser-
vice due to customization-responsiveness challenges
(Swaminathan and Tayur 1998); and double marginal-
ization that leads to lower pro?ts for the supply chain
(Cachon 2003).
The science related to supply chain management
traces its history back to the early 1950s when sev-
eral researchers were interested in understanding
the optimal policies related to inventory manage-
ment. In one of the ?rst works, Clark and Scarf
(1960) developed models for managing inventories at
multiple echelons. Numerous researchers have stud-
ied related inventory problems under stochastic and
deterministic environments over the last 50 years.
This stream of research (mostly from a centralized
perspective) is concisely captured in the research
handbook edited by Graves et al. (1993). Over the
1388 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
last decade, researchers have studied several other
issues related to supply chain management that
include decentralized multiagent models to analyze
supply chain coordination problems, models that inte-
grate information availability across the supply chain
with logistics decisions, models for supply contracts
and demand forecasting, and models that integrate
product design with supply chain management. A
collection of prominent research in this area is con-
tained in Tayur et al. (1999) and Graves and de Kok
(2003).
1.2. E-Business and Supply Chain
E-business can be loosely de?ned as a business pro-
cess that uses the Internet or other electronic medium
as a channel to complete business transactions.
As classi?ed by Geoffrion and Krishnan (2001),
e-business consists of three areas: (1) consumer-
oriented activity and (2) business-oriented activity
supported by (3) the e-business technology infras-
tructure. The consumer-oriented activities consist
of business-to-consumer, consumer-to-consumer, and
government-to-consumer activities. The business-
oriented activities comprise business-to-business,
business-to-government, and government-to-business
activities. The technology infrastructure relates to net-
work infrastructure, network applications, decision
technologies, and software tools and applications.
Within this broad de?nition of e-business activities,
we will restrict our attention mainly to consumer-
oriented and business-oriented activities, as well as
decision technologies that are employed for supply
chain management.
The Internet has in?uenced the usage of supply
chain models in three ways. First, the Internet has
facilitated increased use of enterprise resource plan-
ning (ERP) and advanced planning and optimization
solutions (APS). Second, the ability to obtain real-
time information and the access to large computer
systems is enabling ?rms to develop detailed (high-
granularity) supply chain models that can be uti-
lized to make real-time decisions. Last, the Internet
has created opportunities to integrate information
and decision making across different functional units,
thereby creating a need for supply chain models that
go beyond a business unit to study the extended
enterprise. This has elevated the role of supply chain
models from being decision-making enablers for a
single business unit to being enablers for driving cor-
porate strategy. Thus, the Internet has greatly elevated
the role of supply chain modeling and analysis within
a ?rm.
The advent of e-business has also created sev-
eral challenges and opportunities in the supply chain
environment. First and foremost, the Internet has
increased the opportunity for consumers to buy prod-
ucts and services without going to a store. Though
the practice of direct selling through catalogs and
phone was in use earlier by a few ?rms, the Inter-
net has made this form of sales more signi?cant.
In a direct sales environment, the ful?llment process
determines how long customers will wait between
sale and delivery. This has made the back-end ful-
?llment process—which mostly depends on supply
chain management—extremely important. Further, in
the electronic environment, customer expectations
in terms of quick and timely delivery have also
increased. At the same time, the Internet has opened
up opportunities for ?rms to share information and
ef?ciently coordinate their activities with other enti-
ties in the supply chain. This has created several
new avenues in traditional supply chain areas. For
example, in supplier selection and procurement, ?rms
have to decide if they should join private or pub-
lic exchanges or develop highly-integrated supply
partnerships. They need to determine if they should
use auction and bidding for contracts and, if so,
which type would be most bene?cial. In distribu-
tion, they need to decide if the ?rm will offer prod-
ucts through the Internet channel and, if so, how this
method would differ from the traditional channel.
This raises the question of how the synergies would
be realized in terms of inventory, transportation, and
distribution. Similarly, the availability of real-time
information has raised important questions such as
the degree to which information sharing protocol
should be standard or proprietary; the amount and
type of information that should be shared with the
rest of the supply chain partners; and the types of
collaborative processes that may be bene?cial. The
degree of change in issues related to the supply chain
spans a huge spectrum from concepts and issues that
Management Science/Vol. 49, No. 10, October 2003 1389
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
have been marginally affected to a whole set of new
issues that have emerged as a result of e-business.
First, several issues related to supply chain man-
agement have not necessarily changed in principle,
although e-business may have had an impact on some
of their parameters. For example, to maintain given
levels of service, a ?rm still needs buffer inventory
or buffer capacity. This has not changed as a result
of the Internet, although the uncertainty involved in
the decision making may have decreased with the
availability of more information. Similarly, a ?rm still
needs to take into account the interplay between ?xed
and variable costs, while making decisions related to
procurement or setting up additional capacity. With
the prevalence of the Internet, the ?rm might more
easily be able to obtain a lower procurement price or
salvage excess capacity through market mechanisms.
Next are existing supply chain issues that have
become important as a result of e-business. For exam-
ple, leveraging risk-pooling concepts can greatly ben-
e?t Internet channels because products may be stored
at fewer locations as compared to a traditional dis-
tribution channel. Amazon.com can store all inven-
tory for the entire U.S. market in ?ve warehouses
as opposed to several hundred retail outlets (hence,
stocking points) that would be needed for similar cov-
erage in the traditional channel. Similarly, mass cus-
tomization has gained a lot of momentum with the
Internet because ?rms can allow customers to interac-
tively specify customizations of their offerings. It has
become more important for ?rms to understand how
to cope with customization in an effective manner.
Finally, in the last few years, a third category of
issues new to supply chain management has emerged.
One example is linking the dynamic pricing of prod-
ucts to the inventory and capacity decisions. Another
is coordinating Internet and traditional distribution
channels in terms of prices as well as information
and product ?ows. Additionally, the advent of elec-
tronic marketplaces and auctions has opened a whole
new set of issues related to procurement and supplier
relationships.
1.3. Focus of this Paper
In this paper, we focus on the last two types of sup-
ply chain issues previously discussed: (1) those that
have increased in importance with the Internet, and
(2) those that are new issues in the e-business envi-
ronment. This paper is not intended to be exhaustive
in coverage, rather our aim is to concentrate on spe-
ci?c areas related to supply chain and e-business and
highlight papers that, in our opinion, have tackled
important issues. For a more exhaustive coverage, the
reader is referred to the research handbook by Simchi-
Levi et al. (2003). In particular, our focus will be on
papers that utilize analytical modeling and operations
research methodology in supply chain planning and
execution. This is a relatively new area, thus, many
of the research papers discussed in this article are
unpublished.
1.3.1. Procurement and Supplier Management.
The e-business paradigm has created an immense
opportunity for ?rms to consolidate their buying
processes (also called e-procurement). The ?rst suc-
cess stories in the area of e-procurement came from
software ?rms Ariba (http://www.ariba.com) and
Commerce One (http://www.commerceone.com),
which primarily dealt with indirect goods and
services. The initial idea in these systems was that
when procurement is consolidated under a single
roof, the ?rm bene?ts from price reductions (a direct
result of quantity discount). E-procurement systems
enable individual employees to become aware of
all quali?ed suppliers and to complete the procure-
ment process quickly and easily. Subsequently, more
advanced models for procurement have evolved that
use auctions to determine which ?rm wins a pro-
curement contract. Each contract is open to bids in
an electronic auction house and quali?ed suppliers
are allowed to participate. Although auctions have
been used for industrial contracts before, execution
was rather cumbersome and, thus, were used only
for large and important contracts. E-business ?rms
such as FreeMarkets (http://www.freemarkets.com, a
business-to-business, third-party auctions ?rm) have
made it much easier to conduct these auctions and,
as a result, ?rms are beginning to use these auctions
more frequently for procurement. This poses new
research issues such as whether a ?rm should use
such auctions for all its components and suppliers;
what the long-term effect of such auctions on supplier
relationship is; how ?rms can use both traditional
1390 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
and auction mechanisms to hedge against risks in a
more ef?cient manner; and how a third party such as
FreeMarkets can ensure that capacities of suppliers
are taken into account in these auctions so that the
contract is executable.
A related phenomenon in the supplier management
area is the formation of industry-wide supply chain
consortia such as e2open (high tech) and Covisint
(automobile). The motivation behind the formation
of these consortia (also sometimes called market-
places) is their ability to provide liquidity to inventory
and capacity present in the extended supply chain
and make it possible for the supply chain partners
to get involved in long-term collaborative planning
and design efforts. For example, a manufacturer with
excess inventory could salvage it in the marketplace.
Similarly, buyers may have the ability to conduct auc-
tions for industrial parts and procure capacity options
from the supply base. There are a number of new
research issues that have evolved as a result of the
above changes. For example, how could one quan-
tify the bene?ts of joining such a consortium for a
?rm; how many ?rms of the same type are likely to
stay in a consortium; how could the different ?rms
use the liquidity and options in the marketplace to
improve their operational performance. We will dis-
cuss research on the issues of procurement and sup-
plier management in §2.
1.3.2. Visibility and Information Sharing. The
prevalence of ERP allows ?rms to have access to
data across their supply chains, which could be used
for gaining better ef?ciency and effectiveness (Sodhi
2001). The ability to access information from various
parts of the organization has helped ?rms to stream-
line their business processes and reduce inef?cien-
cies. Although ERP systems were implemented before
the boom in e-business, their potential could not be
explored and expanded due to lack of common stan-
dards and cost of access. The growth of e-business
allows and requires that the information made avail-
able from the ERP systems be shared with other ?rms
in the extended supply chain through the Internet.
This enables ?rms to coordinate and collaborate with
their suppliers and customers as well as synchronize
their in-house operations.
The ability to access information across the supply
chain and use it in real time provides various oppor-
tunities. Inventory requirements for buffer stocks are
likely to be lower, because the uncertainty in forecasts
and demand can be reduced across the supply chain.
Allocation of inventory to different retail outlets or
customers as part of order ful?llment can be done
more effectively when there is visibility about the
number and type of inventory located at the different
sites in the supply chain. As more supply chain exe-
cution information becomes available, ?rms can plan
for future operations using advanced planning and
optimization tools. The ability to share information
creates an opportunity for ?rms to have collaborative
planning and design, which removes the inef?cien-
cies in these processes. This has opened up several
new issues for researchers. For example, if forecasts
are shared more often than before, how should they
be used more effectively; how can collaborative fore-
casts be generated and used across the channel; how
can a ?rm use the demand, inventory, or supply infor-
mation from other partners more effectively. We will
discuss research on these issues in §3.
1.3.3. Pricing and Distribution. Physical retail
stores have traditionally been the most common mode
of distribution. Catalog ?rms have traditionally been
considered a rather niche market for only certain types
of products. E-business has opened up a completely
new dimension to distribution by enabling ?rms to
use alternative distribution channels in addition to
brick-and-mortar stores. Also, the ability to share
information across the supply chain has made it eas-
ier for ?rms to try to coordinate the ?ow of materi-
als across multiple channels. In a rush to exploit this
opportunity, several ?rms during the last few years
have set up Internet stores and created new busi-
nesses (such as online grocery). However, many of
them paid little or no attention to logistics associated
with those operations and, as a result, did not go far.
For example, online grocers grossly underestimated
the transportation costs associated with the “last
mile” delivery and the effect of capacity utilization on
economies of scale (Reinhardt 2001). Traditional ?rms
such as IBM, Barnes & Noble, and Wal-Mart also
faced challenging issues during this period when they
created their own Internet operations. These questions
Management Science/Vol. 49, No. 10, October 2003 1391
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
included how the Internet operation should differ
from traditional practice; how the products should be
priced across these two channels; whether all prod-
ucts should be offered on both channels; and what
kind of autonomy should be provided to Internet
operations.
The ability to dynamically change prices in the
marketplace is yet another aspect of distribution that
e-business has greatly impacted. In traditional sup-
ply chain operations, price changes were allowed
but would occur only at regular or planned inter-
vals. However, the Internet enables ?rms to more
dynamically adjust prices with little additional effort.
This poses interesting issues related to how often
prices should be changed and how one can effectively
couple dynamic pricing with production or capacity
decisions.
The Internet has compounded one of the bene?ts
of distribution that relates to risk pooling. Because
products may be stored at fewer physical loca-
tions, the bene?ts of risk pooling will lead to lower
inventory requirements. Having a separate inven-
tory depot for Internet operations provides the abil-
ity to adjust inventory allocations between the retailer
and the inventory depot. This raises several impor-
tant research issues. How much inventory should be
stored at the various locations? How can ?rms cre-
ate incentives so that the brick-and-mortar retailers
can collaborate with each other? We discuss pertinent
research related to distribution and pricing in §4.
1.3.4. Customization and Postponement. The In-
ternet has increased the expectation of customers for
complete customization at a nominal charge. Even
before the advent of e-business, ?rms faced the chal-
lenges related to mass customization and high prod-
uct variety, but this has increased immensely over the
last few years. Part of the reason is the Internet pro-
vides an easy and convenient way for customers to
express their preferences, and sometimes even view
the product as it is customized. Firms are ?nding
that it is important not to completely rely on tradi-
tional manufacturing paradigms of inventory build-
ing (make to stock). They are ?ne tuning their produc-
tion process so that they can store inventory of raw
materials or semi?nished inventory and more rapidly
respond to customer demand after the customer has
ordered the product. We will discuss several mod-
els that capture the trade-offs in these operations and
their effect on supply chain performance in §5.
1.3.5. Enterprise Software and Decision Support
Technologies. Last, the Internet has had a great
impact on development of decision technologies that
use the available data across the supply chain. In the
past, although the models were available, their imple-
mentation in real time was almost infeasible. As a
result, they were mainly used for planning purposes.
However, the Internet has facilitated real-time access
to information across the supply chain, making it now
possible to use decision models during execution. A
related phenomenon is the development of intelligent
supply chain software agents that could take appro-
priate actions in real time, thus, streamlining supply
chain operations. In §6, we discuss some of these deci-
sion technologies in greater detail.
2. Procurement and Supplier
Selection
One of the major effects of the Internet on supply
chain practices is in the area of procurement. Firms
now use the Internet, not only to diversify the supply
base and hedge risk, but also to obtain lower costs
through auctions. This has raised important issues
related to supply chain management. At the strategic
level, ?rms need to decide whether they should have
long-term contracts with a few ?xed suppliers or use
auctions and a dynamic supplier base to reduce their
costs. In particular, ?rms need to understand under
what circumstances is it bene?cial to have (1) long-
term relationships; (2) auction-based, short-term rela-
tionships; or (3) a combination of (1) and (2). Another
important decision is whether a ?rm should have one
supplier or multiple suppliers and how that choice
may depend on repetition in purchase. Elmaghraby
(2000) presents an extensive survey of economics-
oriented models that deals with supplier selection and
sourcing strategies in traditional settings. Our focus
will be on supplier selection and procurement models
under uncertainty in demand or supply that incorpo-
rate the ability of the ?rm to use the Internet to make
the process more ef?cient.
1392 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
Peleg et al. (2002) develop a model that considers
the potential discounts that an exclusive long-term
supplier may be able to offer based on learning and
compare that to the savings that might be gener-
ated through an auction, in a two-period setting with
uncertain demand. The strategic partner guarantees
a price of p per unit in the ?rst period and p(1 ?A)
in the second period, where 0 - A - 1. This model
attributes it to the long-term learning effects of an
exclusive supplier. In an auction-based setting, the
?rm still pays a per unit price p in the ?rst period, but
conducts an auction in the second period. The ?rm
only knows the distribution of the minimum price
(which is assumed to be independent of the quan-
tity bought) that could be obtained as a result of the
auction. In the combined strategy, the ?rm decides
to use the exclusive supplier for the most part (by
having a contract to order a minimum amount in
the second period) and orders any remaining amount
from the auction. Under random demand assump-
tions, Peleg et al. (2002) derive the optimal ordering
quantities for the ?rm. They show that there exists a
A beyond which it is optimal for the buyer to prefer a
strategic long-term partnership, as opposed to using
an auction in the second period. Further, depending
on the distribution of price obtained from the auction,
the combined strategy may be superior or inferior to
a pure auction strategy. Finally, the authors explore
the effect of increasing the supplier base (in the auc-
tion) by modeling the bene?ts obtained (due to poten-
tial lower prices) in the distribution of price from the
auction.
In the automotive supply chain, a tier-1 manu-
facturer such as Ingersoll-Rand often has to reserve
capacity from its supplier, worldwidepm.com, which
is a coalition of small powder metal technology man-
ufacturers. Because the coalition is a powerful entity,
the manufacturer may need to develop contracting
mechanisms such as capacity reservation (which is
similar in spirit to the minimum order assumptions
of Peleg et al. 2002). Motivated by the above setting,
Erhun et al. (2001) study the strategic interaction and
questions of participation (from the suppliers), pro?t
sharing, and impact on end customers. They ana-
lyze a two-period model where the downward slop-
ing demand in the second period could be in one of
two states. This uncertainty is resolved in the second
period where demand becomes deterministic. In con-
trast to standard two-period models, they assume that
demand only occurs in the second period. The ?rst
period is only for capacity reservation and demand
resolution. Under this setting, they demonstrate the
existence of a unique subgame perfect Nash equi-
librium under the various models. They show that
the supplier, manufacturer, and the supply chain as a
whole are aided by this additional negotiation period
where demand is actually not realized, but demand
resolution occurs and capacity is reserved. Under
uncertain demand and limited capacity at the sup-
plier, they show that the additional period is bene-
?cial, but the impact on pro?ts (and its sharing) is
more involved and depends on the level of capacity
available.
As more and more ?rms begin to use auctions for
conducting business transactions, it is important that
models be developed that capture the buyer bidder
interactions and tie them to supply chain decisions
and constraints. Motivated by FreeMarkets, Gallien
and Wein (2000) study the design of smart markets
for industrial procurement. They study a multi-item
procurement auction mechanism for supply environ-
ments with capacity constraints. This enables them to
model rational behavior of suppliers in terms of their
responses when they have limited capacity. In partic-
ular, they consider an auction with n suppliers and
m components, each having a ?xed order size. The
objective of the manufacturing ?rm is to minimize the
total cost of procurement across all the components,
taking into account the capacities of the various sup-
pliers. In each round, suppliers give a price bid for
each of the components. The assumption is that the
price bids decrease as the rounds progress. Further, at
the end of each round, the auctioneer provides each
of the suppliers information about allocations and the
best bids for the next round that would maximize the
supplier’s pro?ts, assuming all other suppliers hold
on to their old bid. The auction stops when the new
set of bids from all the suppliers is identical to the
previous round. Under these conditions, the authors
show that such an auction does converge, and they
provide a bound for the pro?ts of the manufacturer.
Management Science/Vol. 49, No. 10, October 2003 1393
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
Further, for special cases with two suppliers, they pro-
vide insights on the impact of initial bids.
Davenport and Kalagnanam (2001) study auction
strategies for sourcing large quantities of direct mate-
rials over the Internet. They study two auction mech-
anisms, the ?rst of which solicits supply curves as
bids to capture quantity discounts. The second auc-
tion aggregates short-term demand across multiple
manufacturing facilities and allows suppliers to pro-
vide bundled all-or-nothing bids. They incorporate
business rules as side constraints in their integer pro-
gramming formulation and use computational results
to study these mechanisms.
While an auction typically serves as a price-
determination mechanism, Jin and Wu (2001a) show
that the auction could also serve as a coordina-
tion mechanism for the supply chain. They demon-
strate that different forms of auction and market
mechanisms change the nature of supplier competi-
tion, thus, the buyer-supplier interaction. In partic-
ular, they consider a two-supplier–one-buyer system
under four different types of market schemes. They
propose a two-part contract auction where the buyer
announces a price-sensitive order function, while the
suppliers compete in an ascending bid side-payment
auction. Channel coordination can be achieved if the
market intermediary strives to reduce the extent of
information asymmetry, while restricting the buyer’s
pro?t on the side payments. Using the insights from
the two-supplier–one-buyer analysis, they rank mar-
ket schemes by their impact on expected channel ef?-
ciency, expected pro?tability for the buyer, expected
pro?tability for the winning supplier, and expected
commission revenue for the market maker.
Lee and Whang (2002) investigate the impact of a
secondary market where resellers can buy and sell
excess inventories. They develop a two-period model
with a single manufacturer and many resellers. The
resellers order products from the manufacturer in
the ?rst period and are allowed to trade invento-
ries among themselves in the second period. They
derive the optimal decisions for the resellers, along
with the equilibrium market price of the secondary
market. Two types of effects are created due to the
secondary market: (1) a quantity effect (sales by the
manufacturer), and (2) an allocation effect (supply
chain performance). They show that the former is
indeterminate, i.e., the total sales volume for the man-
ufacturer may increase or decrease, depending on the
critical fractile. However, the latter is always posi-
tive, i.e., the secondary market always improves allo-
cation ef?ciency. The sum of the effects is also unclear
in that the welfare of the supply chain may or may
not increase as a result of the secondary market.
Finally, the authors present potential strategies for
the manufacturer to increase sales in the presence of
the secondary market. Dong and Durbin (2001) study
other implications of such surplus markets on supply
chain ef?ciency. Keskinocak and Tayur (2001) describe
examples where matching algorithms developed by
operations researchers have been used in market-
places as decision support for matching demand and
supply. Kleindorfer and Wu (2003) present a detailed
analysis of research related to integrating long-term
and short-term contracting using B2B exchanges.
A related phenomenon in the area of procurement
and supplier relationships has been industry-wide
consortia where multiple buyers and suppliers within
an industry join and conduct business. Better transac-
tional ef?ciency has been highlighted as the key ben-
e?t of consortia such as Covisint (automotive) and
Converge (high tech). The dynamics of these entities
are not well understood and pose several important
questions. For example, one could expect that having
multiple suppliers on the same platform is likely to
reduce prices for the buyer, but it is likely to bene-
?t other buyers in the consortium as well. Thus, it is
not clear if one should join the consortium in the ?rst
place. One of the reasons for Dell deciding not to join
either of the high-tech consortia, Converge or e2open,
could be that they do not want to open up their
supply chain processes to competitors. Granot and
Sosic (2001) study this issue using a stylized model
where there are three ?rms whose products have a
certain degree of substitutability. They explore con-
ditions under which formation of three-member and
two-member alliances is optimal. In particular, they
consider a deterministic linear model for the demand
that depends on the prices charged by the ?rm and its
competitors, and the degree of substitutability across
the products. They assume that when alliances are
formed, the procurement cost (or the price paid by
1394 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
the ?rm) strictly decreases, and that the sum of the
reductions due to a ?rm joining a two-?rm coalition
with each of the other ?rms is greater than the reduc-
tion obtained when all three ?rms are in the alliance.
With these decreases in costs, the authors derive con-
ditions under which joining a three- or two-party
coalition is bene?cial for a ?rm. They also identify
conditions under which a ?rm may prefer to be in a
two-?rm alliance to prevent others from forming an
independent two-?rm alliance even though this may
lead to lower pro?ts. The authors also provide several
other insights into the stability of these alliances.
Jin and Wu (2001b) study the formation of supplier
coalitions in the context of a buyer-centric procure-
ment market. They consider a second-price descend-
ing sealed-bid auction and propose a two-stage
auction mechanism that allows suppliers to form
coalitions with one another. Building on the founda-
tions of core games and bidding rings, they explore
the idea of managed collusion, which provides a
means to enhancing bidder pro?tability. They also
identify basic requirements for a valid coalition mech-
anism, including characteristics such as individual
rationality, welfare compatibility, maintaining compe-
tition, and ?nancial balance. They show that such a
mechanism could be constructed so that the buyer
does not lose the advantage from supplier competi-
tion, and a stable coalition structure could be formed.
They propose a pro?t distribution scheme among
members in the supplier coalition and show that
the proposed scheme provides proper incentives such
that (1) the best strategy for a coalition member is to
comply with the coalition agreement, and (2) bidding
the true cost is the best strategy so long as the bids are
uniformly distributed and the bidder’s cost is above
a certain threshold. They also investigate the stable
coalition structure under the proposed mechanism,
and show that under symmetric information there
exists one unique strongly stable coalition structure.
Butler et al. (1997) suggest that the prevalence
of the Internet will lead to a reduction in costs of
interaction among ?rms that may make horizontal
alliances (i.e., alliances between similar businesses)
more attractive. Nault and Tyagi (2001) examine
coordination mechanisms that achieve an alignment
between the success of the alliance and its individual
members. They consider coordination mechanisms for
a horizontal alliance characterized by the following
features: (1) ?rms in the alliance can exert effort only
in their local markets to increase customer demand
for the alliance, (2) customers are mobile, and a cus-
tomer living in a given alliance member’s local area
may have a need to buy from some other alliance
member, and (3) the coordination rules followed by
the alliance determine which ?rms from a large pool
of potential member ?rms join the alliance, and how
much effort each ?rm joining the alliance exerts in
its local market. In this horizontal alliance setup,
they consider the use of two coordination mecha-
nisms: (1) a linear transfer of fees between mem-
bers if demand from one member’s local customer is
served by another member, and (2) ownership of an
equal share of the alliance pro?ts generated from a
royalty on each member’s sales. The authors derive
conditions on the distribution of demand external-
ity among alliance members to determine when each
coordination mechanism should be used separately,
and when the mechanisms should be used together.
3. Visibility and Information
Sharing
The Internet has made it easier to share informa-
tion among supply chain partners. The current trend
in the industry is to try to leverage the bene?ts
obtained through information sharing (also called vis-
ibility) across the supply chain to improve opera-
tional performance, customer service, and solution
development. The notion of lack of information in the
supply chain and the resulting bullwhip effect was
?rst studied by Lee et al. (1997). Chen et al. (2000)
quantify the effect of forecasting and lead times on
the bullwhip effect under stylized supply chain set-
tings. A number of other papers have dealt with the
?ow of information in the supply chain and the effect
of that on its performance. We will only focus on a
few papers in the interest of space.
Gavirneni et al. (1999) consider the role of infor-
mation in a two-stage capacitated supply chain under
three types of information ?ow: (1) no information is
shared with the supplier (2) the supplier knows the
end-item demand distribution and the retailer uses
Management Science/Vol. 49, No. 10, October 2003 1395
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
an (s, S) policy, and (3) the supplier has complete
information about the inventory state of the retailer.
The authors show the optimality of order-up-to poli-
cies for ?nite and in?nite horizon models. Through a
computational analysis, they study the role of infor-
mation availability and the savings obtained under
different operational conditions. Lee et al. (2000)
study the value of information in a two-level supply
chain with nonstationary end demand and show that
value of information could be high, particularly in
cases where demand may be correlated over time. The
effect of vendor managed inventory (VMI) systems,
where the buyer shares demand information with the
supplier who, in turn, manages the buyer’s inventory,
have also been extensively studied by researchers
(see Cetinkaya and Lee 2000 and Cheung and Lee
2002).
A related issue in supply chain information shar-
ing is sharing forecast information. It is highlighted
in the popular press that one of the reasons for Dell’s
success is its ability to transmit timely and accurate
forecasts to its suppliers. Because forecasts are not
always accurate, ?rms may pass demand information
in the form of bands (lower and upper estimates) that
may get more re?ned as one gets closer to the actual
period of ordering. It is important to understand
how frequent forecasts can be used to improve sup-
ply chain performance. Kaminsky and Swaminathan
(2001) present a model for forecast evolution that cap-
tures two notions related to forecasts: (1) forecasts are
not exact and (2) forecasts over longer horizons are
less certain than those over shorter horizons. With this
forecast evolution model, they develop a capacitated
production planning model for a single product with
terminal demand. After showing that the optimal pro-
duction policy is order up to, they develop heuristics
for the problem and characterize their performance.
Through a detailed computational study, the authors
show that these heuristics get close to the optimal
solution with holding costs (less than 0.5% away from
optimal on average under the conditions studied)
and provide insights on the effect of early, interme-
diate, and late information updates in forecasts on
optimal costs. Although it is reasonable to assume
that forecasts do get better as one gets closer to the
decision epoch in most cases, Cattani and Hausman
(2000) demonstrate situations when forecast updates
may not get better over time. Miyoaka and Hausman
(2001) study a two-stage supply chain where using
old forecasts may be better.
Although sharing forecasts can be bene?cial in most
cases, and the Internet does facilitate the sharing of
such information, con?icting objectives may cause
the forecasts to be distorted. Celikbas et al. (1999)
address this issue in a intra?rm setting where mar-
keting and manufacturing units may have con?ict-
ing objectives regarding forecasts. Marketing, being
a revenue center, may have an incentive to over-
forecast whereas manufacturing, being a cost center,
may have an incentive to underproduce as compared
to the forecast. The authors study a decentralized
setting where demand is uncertain and marketing
provides a forecast to manufacturing, which in turn
produces a quantity based on the forecasted amount
and the knowledge about the demand distribution.
They show that by suitably setting penalties for over-
forecast and underproduction, one can coordinate the
system. Chen (1999) considers a supply chain whose
members are divisions of the same ?rm and are man-
aged by different individuals who have local inven-
tory information. He shows that the owner of the
?rm can manage the divisions as cost centers without
compromising the systemwide performance by using
an incentive-compatible measurement scheme based
on accounting inventory levels. Similar to Celikbas
et al. (1999), Chen (1999) shows that it is important
for the upstream members of the supply chain to
have access to accurate customer demand informa-
tion. More recently, Cachon and Lariviere (2001) study
forecast sharing in a two-stage supply chain between
a manufacturer and a supplier, where the manufac-
turer provides an initial forecast and a contract to
the supplier, who in turn invests to set up capacity
in anticipation of demand. The focus of the study
is on two types of compliances: (1) forced compli-
ance, where the supplier is forced to develop a given
capacity once he accepts a contract from the man-
ufacturer and (2) voluntary compliance, where the
supplier can set the capacity optimally. Cachon and
Lariviere (2001) show that compliance regime plays
an important role in the supply chain performance.
The authors ?nd that it is always in the interest of
1396 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
the manufacturer to truthfully share the forecast with
the supplier, particularly when the demand forecast
is high.
A growing movement in the industry is collabora-
tive planning, forecasting, and replenishment (CPFR),
which tries to reduce the differences that may exist
between the different supply chain players (retailer,
distributor, manufacturer) regarding demand forecast
for a product. In CPFR, the two parties (typically
manufacturer and retailer) jointly collaborate to gen-
erate a forecast and plan for that forecast. Such a
forecast, which uses the information available to all
supply chain players, is likely to make the supply
chain operations more ef?cient because: (1) the fore-
cast is coordinated, and (2) the forecast carries richer
information. Aviv (2001) studies a two-stage two-
player cooperative supply chain. He assumes that the
supply chain is decentralized so that information such
as inventory position and individual forecasts are
only locally available. Under collaborative forecast-
ing, Aviv (2001) assumes that both the manufacturer
and retailer maintain and use a single forecast process
for production replenishment. He compares the ben-
e?ts of collaborative forecasting and local forecasting
in a setting with stationary demand distribution.
4. Distribution and Pricing
Traditional research in the area of distribution has
mainly focused on three issues: (1) the optimal
location of distribution centers and warehouses,
(2) inventory and allocation in a distribution net-
work, and (3) transportation and routing algorithms.
Such models have been studied by researchers in
operations research (OR) for several years. For exam-
ple, Geoffrion and Graves (1974) study the multi-
commodity warehouse location problem for a major
food ?rm; Eppen and Schrage (1981) study the
inventory allocation decisions in a distribution net-
work; and Bramel and Simchi-Levi (1997) present an
overview of logistics research. During these years,
OR researchers have developed effective solution
methodologies for solving large-scale instances of the
production distribution problem that have enabled
?rms to streamline their supply chains. Arntzen et al.
(1995) describe the implementation of such a model
for reengineering at DEC.
The Internet has provided another channel for dis-
tribution of goods. With the prevalence of business-
to-consumer and business-to-business e-commerce,
several ?rms such as Amazon have had to develop
a comprehensive distribution network for their
business. Clearly, the customer base is more dis-
persed and in many cases, customers may buy goods
that need to be delivered elsewhere (such as a gift).
Neglecting the costs and synergies related to distribu-
tion and transportation have proved fatal to several
early movers in this business. In particular, ?rms that
have been most affected by this were online grocers
(like Peapod and Webvan) and home delivery ?rms
(like Kozmo and HomeRuns). This primarily occurred
because these ?rms had to deal with numerous tight
delivery time windows (mostly customer preferred)
in addition to other logistic complexities, which made
it extremely dif?cult to gain ef?ciency through syner-
gies in distribution and transportation.
From a supply chain standpoint, the integration
of traditional and Internet distribution channels is
attractive because it promises pro?t gains, inventory
reduction, and increased customer service. However,
many of the traditional channels are not necessarily
centrally controlled. Seifert et al. (2001) study inven-
tory coordination contracts and their impact on the
direct and indirect channels. Motivated by an actual
business line at Hewlett Packard, they consider a situ-
ation where the ?rm has N traditional identical retail-
ers and one virtual store and random demand. The
assumption is that when a customer is unable to get
the product at the virtual store, then she is willing to
wait. If there is excess inventory at the retailer out-
let at the end of the period, then it is used to satisfy
the demand. Retailers individually make the inven-
tory decisions in a decentralized system, whereas
the manufacturer makes decisions related to inven-
tory at the virtual store. When the channels are not
integrated, then a menu of buy-back contracts can
coordinate the system. In an integrated supply chain
(where excess inventory from the traditional chan-
nel is used to satisfy excess demand from the virtual
store), Seifert et al. (2001) show that when a trans-
fer payment scheme is introduced in addition to a
buy-back scheme, the decentralized system can be
coordinated. Through a detailed computational study,
Management Science/Vol. 49, No. 10, October 2003 1397
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
they provide several managerial insights into pro?ts
and inventory as a function of number of retailers and
degree of coordination.
Anupindi and Bassok (1999) consider a model with
two retailers and one manufacturer, and explore the
bene?ts to the manufacturer and the channel if the
two retailers centralized their inventory in a lost sales
setting. Their study was motivated by some early
pilots by General Motors where the ?rm was explor-
ing how demand and supply availability information
in the distribution channel could be more bene?cially
used. The authors ?nd that centralization of stock
by the retailers need not be in the best interest of
the manufacturer. The true bene?ts to the manufac-
turer may depend on the service levels of the retail-
ers and the degree of substitution (which they term
“search”) of demand between the two retailers. The
authors prove that there exists a threshold level of
search beyond which it is detrimental to manufac-
turer’s pro?ts if the retailers centralize their inven-
tories when the wholesale price is ?xed. For high
search parameters, even the supply chain pro?ts may
decrease as a result of centralization. Even under an
optimal wholesale price setting, they show that a
manufacturer may be better off with a decentralized
system when the degree of search is high.
Netessine and Rudi (2001) study incentive issues
related to drop shipping in business-to-consumer sup-
ply chains. In the drop-shipping method (commonly
used by many e-tailers) the retailer is primarily con-
cerned with customer acquisition (as a result, does not
store any inventory) and the wholesaler takes inven-
tory risk and performs ful?llment. They analyze three
different drop-shipping methods in a multiperiod set-
ting with different power structures: powerful retailer,
powerful wholesaler, and both equally powerful.
These power structures determine the sequence of
actions taken by the ?rms and whether the resulting
equilibrium is Nash or based on Stackelberg. In par-
ticular, they assume that demand is stochastic and the
demand distribution is a linear combination of a ran-
dom component and a function of effort expended
by the retailer for customer acquisition. Netessine
and Rudi (2001) also assume that the mean demand
is increasing and concave in the customer acquisi-
tion spending. They obtain optimal solutions for the
centralized system and traditional (where the retailer
decides both customer acquisition effort and quanti-
ties) and drop-shipping (where the retailer decides on
the effort and the wholesaler decides on the quantity)
methods, and show that these decentralized systems
are not as ef?cient. Further, Netessine and Rudi (2001)
show that traditional contracts based on buy-back
or quantity discounts cannot coordinate the above
decentralized systems. They show that a contract that
also includes a compensation for the retailer’s market-
ing expense can coordinate the system. Finally, several
numerical insights on the problem are provided.
Anupindi et al. (2001) develop a general frame-
work for modeling various issues in decentralized
distribution channels such as drop shipping, ful?ll-
ment houses, and inventory speculators. They con-
sider N retailers who face stochastic demand and hold
stocks locally and/or at one or more central ware-
houses. They assume a given level of substitutability,
so if a retailer is stocked out, a certain portion of
that demand can be satis?ed from existing inven-
tory at other retailers or the central warehouses.
The main decision relates to planning for inventory
and allocation. In a “coopetitive” framework, they
develop the notion of a claim that establishes owner-
ship rights for each unit of inventory in the system
regardless of its location. Thus, by buying a claim on a
unit, a player gets the right to determine how the unit
will be used once demand is realized. This enables
analysis of the inventory decision as a noncooperative
game. Simultaneously, the authors assume that the
shipping and allocation decisions are cooperatively
made. Using the concept of a core (an allocation
scheme, which makes sure that coalitions are stable),
they develop suf?cient conditions for the existence of
one. They also develop conditions for the existence
of a Nash equilibrium for the inventory decisions.
Under the above conditions, they show the existence
of an allocation mechanism that leads to a coordina-
tion across the supply chain and develop conditions
for its uniqueness.
Cattani and Souza (2002) study a model where
an Internet distributor determines the inventory to
stock given that there may be alternative categories of
customers (who may differ in delivery requirements)
and that there may be alternative ways to ship the
1398 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
product. In particular, they assume that there are two
types of customers: (1) those who want air shipment
and (2) those who want ground shipment. Using a
queuing model with Poisson arrivals and exponen-
tial service times, they analyze ?ve different allo-
cation policies. These range from simple ones such
as ?rst-in, ?rst-out (FIFO) to more profound poli-
cies that take into account the inventory position and
prioritize the customer types accordingly. One such
policy serves both types of customers until inven-
tory drops below a threshold S
?
, beyond which only
the air shipments are completed (while the ground
orders may be backlogged or lost). Once the inventory
level increases beyond the threshold (assuming the
machine is always running), the backlogged demand
is upgraded to an air shipment. Using a computa-
tional study, they provide insights into the usefulness
of such ?exibility.
In traditional supply chain research, pricing has
not been an issue because models have been devel-
oped for a single entity whose demand is exogenously
prescribed (of late, there are models for competition
being analyzed; see Cachon 2003). However, with
the advent of the Internet, even a single ?rm needs
to decide how to differentially price products over
the traditional and the Internet channels. Brynjolfsson
and Smith (2000), in their empirical study of book and
CD prices over a 15-month period, concluded that the
Internet prices are 9%–16% lower than prices in con-
ventional outlets. Further, how ?rms price their prod-
ucts on the traditional and the Internet channels may
depend, to a great extent, on the ?rm’s strategy and
the degree of autonomy provided to the Internet arm
of the ?rm. For example, some ?rms see the Internet
as a medium of convenience for customers trying,
instead, to provide useful information on their prod-
ucts, but few discounts. On the other hand, several
?rms (like Wal-Mart) spun off their Internet arms as
independent entities and gave them complete free-
dom to operate their own way.
Cattani et al. (2002b) study optimal pricing for a
?rm that sells on both traditional and Internet chan-
nels, given that the cost and customer preferences in
each segment could differ. In particular, they consider
a linear customer utility model where each customer
has an independent preference for the Internet and
traditional channels. The customer’s choice of channel
depends on the prices of both channels and the cus-
tomer’s individual preference. Under the assump-
tion of uniformly distributed customer preferences,
Cattani et al. (2002b) analyze the impact of vari-
ous pricing and governance issues under monopoly
and competition. Under monopoly, they consider
four different pricing schemes: (1) where traditional
channel prices are not changed and the Internet
channel optimizes its own pro?ts, (2) where tradi-
tional channel prices are unchanged but the Internet
channel prices are set to optimize the joint pro?t, (3)
where both channels price identically but the optimal
price is chosen, and (4) where both prices are jointly
optimized. Cattani et al. (2002b) show that optimal
prices depend, to a great extent, on customer pref-
erences and cost of the two channels. Further, they
establish threshold levels in the difference between
the traditional and Internet costs, beyond which it
is optimal to price one channel higher than the
other. In the competitive case, they consider a tradi-
tional monopolist who is competing against a pure
e-tailer. Cattani et al. (2002b) show the existence of a
unique Nash equilibrium. With a detailed computa-
tional study, the authors provide interesting insights
that, under many conditions, a ?rm, in fact, could
be close to optimal under case (2) above where they
do not change the traditional channel but align the
Internet channel with the existing business. Also, the
joint pricing strategy in case (3) is reasonable when
the costs are not different across the two channels.
Finally, Cattani et al. (2002b) show that case (1), a
strategy adopted by many ?rms over the last cou-
ple of years, could be quite detrimental to the ?rm’s
pro?ts. Using a stylized linear downward sloping
demand function, Huang and Swaminathan (2003)
analytically prove that pro?ts under cases (2) and (3)
are no more than 4% away from the optimal. Further,
they show that under competition, a retailer that has
both the Internet and traditional channels, will price
products higher than a pure Internet retailer.
Druehl and Porteus (2001) examine price compe-
tition between two ?rms, one that has an Internet
channel and another with a brick-and-mortar chan-
nel, under similar demand assumptions as above.
Additionally, they assume that only a fraction of the
Management Science/Vol. 49, No. 10, October 2003 1399
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
population has access to the Internet and is willing to
buy online. They solve for the ?rm’s optimal prices
in Nash and Stackelberg games and relate the com-
petitive outcomes (market structure) in terms of the
degree of innovation. The degree of innovation is
effectively a maximum possible margin (maximum
reservation price minus unit cost) ratio for the two
?rms. As the fraction of consumers willing to buy
online increases, they ?nd that the duopoly outcome
becomes more likely. Thus, even when almost every-
one is willing to buy online, both ?rms are likely to
make positive pro?ts. Using the degree of innova-
tion, they discuss the importance of marketing and
cost control. Balasubramanian (1999) studies a related
problem where he analyzes the competition between
a direct channel and traditional retailers. He models
the customer space in the form of a circle where
the customers are uniformly distributed around the
circumference. The direct channel is located at the
center of the circle equidistant from all the cus-
tomers, thereby having a single location or preference
parameter that determines both the traditional and
direct channel distance for a customer (because the
distance from the center is ?xed at the radius of
the circle). Given this model of the customer space,
Balasubramanian (1999) derives the competitive equi-
librium prices between traditional and direct chan-
nel, and explores the impact of information related to
availability of the product from the direct channel to
customers.
Another related issue is dynamic pricing. With the
advent of the Internet and greater connectivity, ?rms
have a greater ability to dynamically price their prod-
ucts. This approach has been successfully adopted by
?rms in the airline, hotel, and rental car businesses
where the product or service has a perishable nature.
Auctions and bidding are also sometimes referred
to as dynamic pricing models, and have become
prevalent in an e-business environment as discussed
in §3. Another related dynamic pricing model relates
to pricing products based on supply chain charac-
teristics such as of?oading excess capacity. The key
issue is the integration of production and pricing
decisions through the supply chain. Motivated by
an initiative at General Motors, Chan et al. (2001)
study a model that develops dynamic prices, which
are integrated with the production and inventory
control for nonperishable products in a multiperiod
setting. Chan et al. (2001) allow for periodically vary-
ing demand, holding costs, capacities, and produc-
tion costs. In particular, they consider a model where
demand is assumed to be a nonincreasing determin-
istic function of price and the revenue function is
assumed to be concave. Further, they restrict both
production and inventory to be integral values and
have upper and lower bounds on the price the ?rm
can charge in each period. With these assumptions,
they show that the constraint matrix is a polymatroid
and prove that a greedy allocation algorithm for pro-
duction is optimal. Using an extensive computational
study, they offer various interesting insights into
the bene?ts obtained from dynamic pricing under
different operational conditions related to capacity,
demand, and frequency of price changes.
5. Customization and
Postponement
Traditional inventory management models assume
that inventory is stocked and demand is satis?ed
based on availability (called make-to-stock systems).
Swaminathan and Tayur (2003) provide an exten-
sive review of research advances in this area. For
most consumer products, the make-to-stock system
was an appropriate business model. However, the
Internet has resulted in a growing movement toward
selling directly to customers, which has increased
the degree of customization. For manufactured prod-
ucts, ?rms have moved to a build-to-order envi-
ronment or have adopted postponement (also called
delayed differentiation). Venkatesh and Swaminathan
(2003) describe successful implementations related
to postponement across industries and discuss costs
and bene?ts associated with alternative types of
postponement. Swaminathan (2001b) describes exam-
ples that illustrate the increase in customization
due to the Internet. For example, ewatchfactory
(http://www.ewatchfactory.com) enables a customer
to completely customize a watch. This business model
depends on the Internet and a postponement strat-
egy at the ?nal assembly. Another ?rm, Timbuk2
1400 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
(http://www.timbuk2.com), allows consumers to cus-
tomize their computer bags, messenger bags, and lap-
top sleeves. Again, such a business opportunity could
be exploited due to the Internet and a clever combi-
nation of make-to-order and make-to-stock strategies.
In the next few passages, we highlight a handful of
papers that deal with postponement and customiza-
tion models. For a detailed overview on postpone-
ment models, see Swaminathan and Lee (2003).
Lee (1996) describes the most basic version of
the delayed differentiation model where there are
A products and inventory is carried in ?nished form.
All products are customized from the inventory avail-
able at the end of the standard steps of the process.
This model is similar to the one-warehouse multi-
retailer inventory problem studied by Eppen and
Schrage (1981). The basic assumption is that the stan-
dard part of the process takes | time periods to com-
plete, and the remaining T ?| periods correspond to
time for the customization step. This is analogous
to the warehouse lead time of | periods and T ?|
periods of transportation time from the warehouse to
the retailers. All the products are assumed to have
independent and normal demands with mean j
i
per
period and standard deviation u
i
per period. Lee
(1996) analyzes this system with process standard-
ization and addresses the impact of postponement,
which is re?ected in the parameter |. He shows that
the variance of the inventory is decreasing in |. Thus,
postponement will lead to reduction of inventory of
?nished products. Further, the reduction in inven-
tory is greater when the end product demands are
negatively correlated. Lee (1996) also shows that the
reduction in variance is greater when the number
of identical end products is larger. Lee and Whang
(1998) further explore this model by assuming that
demands are not independent and identically dis-
tributed (IID) over time. With non-IID demand, the
value of postponement is more than just the ability to
make product commitments at a later point in time
when realized demands have been revealed. The pro-
gression of demands may also help to improve the
forecast of the future demands.
The above models share an implicit assumption
that the production-distribution process is contin-
uous, and inventory can be stored only in ?n-
ished products form. In general, manufacturing
environments involve a discrete set of operations,
and inventory can be stored immediately following
any one of these stages. Furthermore, the costs asso-
ciated with delaying differentiation have not been
considered in these. Lee and Tang (1997) consider
a model where two products require N sequential
tasks for completion where inventory can be stored
in a buffer after each task, with the buffer after the
Nth task being ?nished goods. The ?rst | tasks are
assumed to have been standardized, i.e., the inven-
tory in the buffer after the |th operation can be used
for customization into either product. The tasks | +1
to N are distinct for the two ?nal products. Thus,
the point of differentiation is right after the |th step.
Under a normal demand assumption for the two
products and a discrete time setting, Lee and Tang
(1997) consider the costs associated with standardiz-
ing any stage of task and develop interesting insights.
Garg and Tang (1997) extend the analysis to a system
with two points of differentiation: the family differen-
tiation, and the product differentiation points.
The above papers assume that the production
distribution process does not have any capacity
constraints. Swaminathan and Tayur (1998) analyze
a ?nal assembly process with production capacity
where inventory is stored in an intermediate form
(vanilla boxes). In addition to the intermediate form,
they allow the two extreme forms of vanilla boxes:
as components, and as ?nished products. Therefore,
this model captures both assemble to order (where
components are stocked and products are assem-
bled from the components after demand is realized)
and make to stock (where inventory is carried in
?nished form only) as special cases. This approach
allows for multiple points of differentiation, as there
is no restriction on the type of vanilla box stored.
The ?nal product demands are assumed to have
any stochastic distribution allowing for correlation
across product demands. Swaminathan and Tayur
(1998) assume that the vanilla box inventory follows
a base stock policy. The inventory is brought up
to that level in every period, and when demand is
realized, products are assembled from vanilla boxes
by adding other components within the production
capacity. Unsatis?ed demand is lost with a penalty
and the remaining inventory of vanilla boxes incurs
Management Science/Vol. 49, No. 10, October 2003 1401
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
holding cost. Using a stochastic integer program and
an ef?cient simulation-based algorithm, Swaminathan
and Tayur (1998) explore the bene?ts of postpone-
ment through vanilla boxes under various settings.
Among other results, they show that postponement
using vanilla boxes outperforms both assemble-to-
order and make-to-stock systems when the assembly
capacity available is neither too slack nor too tight
(representative of most real environments). Further,
they ?nd that the vanilla box approach is extremely
powerful under high variance and negative corre-
lation among product demands. Finally, they pro-
vide examples where stocking two types of vanilla
boxes may be suf?cient for a product family with
10 products, and the performance may be better
than a make-to-stock approach (with all 10 products).
Several other authors since have studied the post-
ponement and the make-to-order versus make-to-
stock problem under various conditions (see Graman
and Magazine 2002, Benjaafar and Gupta 2000).
Rajagopalan (2002) studies a multiitem problem
with setup times between production, limited capac-
ity, and congestion effects. In such an environment,
making an item to order reduces inventory but may
increase the lot size and inventory costs for other
items that are made to stock. Further, lead times
increase due to congestion effects, which may lead to
higher safety stocks for made-to-stock items and poor
service for made-to-order items. Given this tradeoff,
Rajagopalan (2002) develops a nonlinear integer opti-
mization model to determine which products should
be make to order and which should be make to stock
to minimize the total costs, and develops an ef?cient
heuristic for the problem. More recently, Cattani et al.
(2002a), motivated by customization and capacity
issues faced by Timbuk2, develop a two-stage frame-
work involving marketing and operations models to
determine the conditions under which a ?rm bene-
?ts from using ?exible capacity to produce custom
products as demanded each period, and then ?lling
the production schedule with make-to-stock output of
standard products to restock inventory. This mitigates
the “lumpy” demand for custom products, improv-
ing capacity utilization as compared to a focused
approach where standard items are made with ef?-
cient capacity and custom products with ?exible
capacity. Cattani et al. (2002a) explore how capacity
decisions impact demand parameters, and pricing
impacts capacity decisions and, ultimately, production
cost. The authors provide conditions under which
convergence of the two models is achieved, and illus-
trate their framework with data from the messenger
bag manufacturer.
Clearly, enabling customization in the product line
is a great challenge to ?rms. Although the Internet
facilitates a business model around customization,
Zipkin (2001) describes the limits of mass customiza-
tion. Swaminathan (2001b) presents a managerial
framework based on standardizing operations that
could potentially mitigate the negative effects of
customization.
6. Decision Technology
One of the biggest impacts of the Internet on sup-
ply chain practices has been the development of dif-
ferent types of decision technologies, which can be
used in supply chain planning and execution. This
is also a result of the availability of accurate infor-
mation that could increase the usefulness of deci-
sion models to a great extent. i2 (http://www.i2.com)
and Manugistics (http://www.manugistics.com), two
of the leading ?rms in the supply chain planning
software industry, were the ?rst to develop constraint-
based planning approaches (linear and nonlinear pro-
grams) that tapped into data available in ERP systems
to provide better supply chain planning and execu-
tion decisions as compared to earlier manufacturing
resources planning (MRP) systems. With the growth
in e-business, these and several other ?rms such as
Agile, Inc. (http://www.agile.com) have developed
more solutions that leverage the Internet to share
information, and have created a suite of supply chain
solutions. Academics in the supply chain area have
also played an active role in creating newer decision
technologies that use models developed in research,
and have suf?ciently adapted them to be used in
practice (see Table 1).
Clearly, better decision technologies lead to better
and more ef?cient decisions. However, the success of
the implementation of decision technology depends,
to a great extent, on how it is eventually used in
1402 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
Table 1 A Sample of Firms That Have Used Theoretical Advancements
in Logistics and Supply Chain Management in Their Software
Solutions
Company Description
Evant Uses results from stochastic inventory theory
(www.nonstop.com) and daily forecasts to set inventory level
at the retail sites
SmartOps Provides decision support for supply chain
(www.smartops.com) planning (tactical and strategic) and
execution by tapping into real-time data and
using algorithms to determine supply chain
decisions, such as total chain inventory
levels, joint replenishment, and lot sizes
Manhattan Associates Provides a suite of tools that enables real-time
(www.logistics.com) sharing of information about transportation
capacities and helps in reducing inef?ciencies
AspenTech Provides ?rms in the process industry with
(www.aspentech.com) real-time information and analytical and
simulation tools to optimize their most
important processes
4R Systems Provides decision support to forecast and plan
(www.4rsystems.com) for production and inventory management
more ef?ciently in the retail sector
Optiant, Inc. Provides software to help manufacturers
(www.optiant.com) design and con?gure their supply chains
based on inventory and responsiveness
needs
LogicTools Provides optimization-based decision-support
(www.logic-tools.com) tools for logistics and inventory management
MCA Solutions Provides real-time information and decision
(www.mcasolutions.com) support for managing service parts inventory
the business process. Palmer and Markus (2000),
through an empirical study, explore the interrelation-
ship between business process change (in the context
of quick response in the retail industry), degree of
new information technology, and ?rm performance.
Decision technologies based on real-time agents
for supply chain management that use optimization
theory, agent behavior, and economic theory may
become important in the future. Wellman (1993) con-
siders a market-oriented programming approach to
distributed problem solving by deriving the activi-
ties and resource allocations for a set of computa-
tional agents. He considers the multicommodity ?ow
problem and shows that careful construction of the
decision process, according to economic principles,
can lead to ef?cient distributed resource allocation,
and that the behavior of the system can be meaning-
fully analyzed in economic terms. Swaminathan et al.
(1998) propose the concept of intelligent agents in the
supply chain that may have local and global informa-
tion about the world around them and take appro-
priate actions (using decision models), depending on
the situation. They provide a framework of agents
and decision rules. With the Internet, it is possible
to create such software agents that may be able to
tap into required information from other agents and
entities in the supply chain. More recently, Fan et al.
(2003) discuss the design principle of a decentral-
ized supply chain organization that uses the auction
market as the coordinating mechanism. They present
a bundle-trading market that can auction multiple
resources in bundles. They show that such a system
could provide the right incentive for all the partici-
pants to act in the best interest of the supply chain.
This will take supply chain decision technology to
the next level where distributed decision making may
become more prevalent.
7. Concluding Remarks
In this paper, we have provided an overview of
relevant analytical models in the area of e-business
and supply chain management. The topic is broad,
so we restricted our attention to areas in the sup-
ply chain, which have been signi?cantly impacted
by the advent of the Internet. We concentrated our
efforts on a few papers that, in our opinion, have
addressed important issues. Our focus has mainly
been on research in operations management, opera-
tions research, and management science. We recog-
nize that papers in marketing, information systems,
economics, and other areas address other important
issues related to electronic supply chains.
As ?rms strive to adapt their supply chains to
the e-business environment, they need to decide how
much and in what sequence they will be investing
in various supply chain efforts to leverage the ben-
e?ts of the new technology. Traditional models have
focused on ?nding the optimal strategy for a single
parameter (such as inventory, forecast, lead time,
capacity, price). However, there is a growing need
for models that can provide insights into the relative
impact of altering the different parameters in the sup-
ply chain. For example, a ?rm with limited resources
Management Science/Vol. 49, No. 10, October 2003 1403
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
may need to know whether it is better to invest in
vendor-managed inventory or reduce the lead time
with existing suppliers. Similarly, research that helps
?rms identify supply chain changes that have quick
bene?ts versus longer term results will be in demand.
As more and more ?rms begin to integrate their
online and traditional operations and share more
information over the Internet, real-time supply chain
management models on product life-cycle manage-
ment, dynamic pricing and production coordination,
integrated models for supply consortia, and the coor-
dination of Internet and traditional channels, are
going to become all the more signi?cant.
Acknowledgments
The authors thank the editors, an anonymous associate editor, and
four anonymous referees who provided valuable input and com-
ments that have contributed to improving the content and exposi-
tion of this paper.
References
Anupindi, R., Y. Bassok. 1999. Centralization of stocks: Retailers
versus manufacturers. Management Sci. 45(2) 178–191.
, , E. Zemel. 2001. A framework for analysis of decentral-
ized distribution systems with implications for e-commerce.
Manufacturing Service Oper. Management 3(4) 349–368.
Arntzen, B., G. Brown, T. Harrison, L. Trafton. 1995. Global supply
chain management at DEC. Interfaces 25(1) 69–94.
Aviv, Y. 2001. The effect of collaborative forecasting on supply chain
performance. Management Sci. 47(10) 1326–1343.
Balasubramanian, S. 1999. Mail vs. mall: A strategic analysis
of competition between direct marketers and conventional
retailers. Marketing Sci. 17(3) 181–195.
Benjaafar, S., D. Gupta. 2000. Make-to-order, make-to-stock or delay
product differentiation?—A common framework for model-
ing and analysis. Working paper, Department of Mechanical
Engineering, University of Minnesota, Minneapolis, MN.
Bramel, J., D. Simchi-Levi. 1997. The Logic of Logistics. Springer-
Verlag, New York.
Brynjolfsson, E., M. D. Smith. 2000. Frictionless commerce? A com-
parison of Internet and conventional retailers. Management Sci.
46(4) 563–585.
Butler, P., T. W. Hall, A. M. Hana, L. Mendonca, B. Augus-
tine, J. Manylika, A. Sahay. 1997. A revolution in interaction.
McKinsey Quart. 1 5–23.
Bylinski, G. 2001. The e-factory catches on. Fortune (July 23).
Cachon, G. 2003. Supply chain coordination with contracts.
S. Graves, T. de Kok, eds. Supply Chain Management—
Handbook in OR/MS, Vol. 11. North-Holland, Amsterdam, The
Netherlands.
, M. Lariviere. 2001. Contracting to assure supply: How to
share demand forecasts in a supply chain. Management Sci.
47(5) 629–646.
Cattani, K., W. Hausman. 2000. Why are forecast updates often
disappointing? Manufacturing Service Oper. Management 2(2)
119–127.
, G. Souza. 2002. Inventory rationing and shipment ?exi-
bility alternatives for direct market ?rms. Production Oper.
Management 11(4) 441–457.
, E. Dahan, G. Schmidt. 2002a. Spackling: Smoothing make-
to-order production of custom products with make-to-
stock production of standard items. Working paper, Kenan-
Flagler Business School, University of North Carolina, Chapel
Hill, NC.
, W. Gilland, J. M. Swaminathan. 2002b. Coordinated pricing
on traditional and Internet channels. Working paper, Kenan-
Flagler Business School, University of North Carolina, Chapel
Hill, NC.
Celikbas, M., J. G. Shanthikumar, J. M. Swaminathan. 1999. Coor-
dinating production quantities and demand forecasts through
penalty schemes. IIE Trans. 31(3) 851–864.
Cetinkaya, S., C. Y. Lee. 2000. Stock replenishment and shipment
scheduling for vendor managed inventory. Management Sci.
46(2) 217–232.
Chan, L., D. Simchi-Levi, J. Swann. 2001. Dynamic pric-
ing for manufacturing with stochastic demand and dis-
cretionary sales. Working paper, Massachusetts Institute of
Technology, Cambridge, MA.http://www.isye.gatech.edu/
˜jswann/publications/SwannManuDynPricing.pdf.
Chen, F. 1999. Decentralized supply chains subject to information
delays. Management Sci. 45(8) 1076–1090.
, Z. Dresner, J. K. Ryan, D. Simchi-Levi. 2000. Quantifying
the bullwhip effect in a simple supply chain: The impact of
forecasting lead time and information. Management Sci. 46(3)
436–443.
Cheung, K. L., H. L. Lee. 2002. The inventory bene?t of ship-
ment coordination and stock rebalancing in a supply chain.
Management Sci. 48(2) 300–306.
Clark, A., H. Scarf. 1960. Optimal policies for a multi-echelon inven-
tory problem. Management Sci. 6 475–490.
Davenport, A. J., J. R. Kalagnanam. 2001. Price negotiations for
procurement of direct inputs. B. Dietrich, R. Vohra, eds. Math-
ematics of the Internet: Auctions and Markets. Springer-Verlag,
New York.
Dong, L., E. Durbin. 2001. Markets for surplus components with a
strategic supplier. Workingpaper, OlinSchool of Business, Wash-
ington University, St. Louis, MO.http://www.olin.wustl.edu/
faculty/durbin/Papers/DongDurbinSurplusMarketMS.pdf.
Druehl, C., E. Porteus. 2001. Price competition between an Internet
?rm and a bricks and mortar ?rm. Working paper, Graduate
School of Business, Stanford University, Stanford, CA.
Elmaghraby, W. J. 2000. Supply contract competition and sourcing
policies. Manufacturing Service Oper. Management 2(4) 350–371.
1404 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
Eppen, G., L. Schrage. 1981. Centralized ordering policies in a
multi-warehouse system with lead times and random demand.
L. B. Schwartz, ed. Multi-Level Production Inventory Control
Systems: Theory and Practice. North-Holland, Amsterdam, The
Netherlands, 51–58.
Erhun, F., P. Keskinokak, S. Tayur. 2001. Sequential procurement in
a capacitated supply chain facing uncertain demand. Working
paper, Graduate School of Industrial Administration, Carnegie
Mellon University, Pittsburgh, PA.http://www.gsia.cmu.edu/
andrew/stayur/feryal.ps.
Fan, M., J. Stallaert, A. B. Whinston. 2003. Decentralized mechanism
design for supply chain organizations using an auction market.
Inform. Systems Res. 14(1) 1–22.
Fisher, M. L., J. H. Hammond, W. R. Obermeyer, A. Raman. 1994.
Making supply meet demand in an uncertain world. Harvard
Bus. Rev. 72(3) 83–92.
Gallien, J., L. Wein. 2000. Design and analysis of a smart market
for industrial procurement. Working paper, Massachusetts
Institute of Technology, Cambridge, MA.http://web.mit.edu/
lwein/www/smarket.pdf.
Garg, A., C. S. Tang. 1997. On postponement strategies for product
families with multiple points of differentiation. IIE Trans.
29 641–650.
Gavirneni, S., R. Kapuscinski, S. Tayur. 1999. Value of information
in capacitated supply chains. Management Sci. 45 16–24.
Geoffrion, A., G. W. Graves. 1974. Multicommodity distribution
system design using benders decomposition. Management Sci.
20(5) 822–826.
, R. Krishnan. 2001. Prospects for operations research in the
e-business era. Interfaces 31(2) 6–36.
Graman, G. A., M. J. Magazine. 2002. A numerical analysis of
capacitated postponement. Production Oper. Management 11(3)
340–357.
Granot, D., G. Sosic. 2001. Formation of alliances in Internet-based
supply exchanges. Oper. Res. Forthcoming.
Graves, S. C., A. G. de Kok. 2003. Supply Chain Management—
Handbook of OR/MS. North-Holland, Amsterdam, The
Netherlands. Forthcoming.
, R. Kan, P. H. Zipkin. 1993. Logistics of Production and
Inventory—Handbook of OR/MS. North-Holland, Amsterdam,
The Netherlands.
Huang, W., J. M. Swaminathan. 2003. Pricing in mixed channels:
Analysis and bounds. Working paper, Kenan-Flagler Business
School, University of North Carolina, Chapel Hill, NC.
Jin, M., S. D. Wu. 2001a. Supply chain coordination in
electronic markets: Auction and contracting mecha-
nisms. Working paper, Lehigh University, Lehigh, PA,http://www.lehigh.edu/˜sdw1/ jin1.pdf.
, . 2001b. Supply chain coalitions in e-commerce auc-
tions: Validity requirements and pro?t distribution. Working
paper, Lehigh University, Lehigh, PA,http://www.lehigh.edu/
˜sdw1/jin2.pdf.
Johnson, C., K. Delhagen, A. Dash. 2002. US e-commerce: The next
?ve years. Forrester Report (August 27).
Kaminsky, P., J. M. Swaminathan. 2001. Utilizing forecast band
re?nement for capacitated production planning. Manufacturing
Service Oper. Management 3(1) 68–81.
Keenan, F., S. Ante. 2002. The new teamwork. Bus. Week
(February 18).
Keskinocak, P., S. Tayur. 2001. Quantitative analysis of Internet
enabled supply chains. Interfaces 31(2) 70–89.
Kleindorfer, P., D. J. Wu. 2003. Integrating long-term and short-term
contracting via business-to-business exchanges for capital-
intensive industries. Management Sci. 49(11) Forthcoming.
Lee, H. L. 1996. Effective management of inventory and service
through product and process redesign. Oper. Res. 44 151–159.
, C. Billington. 1992. Managing supply chain inventories: Pit-
falls and opportunities. Sloan Management Rev. 33(3) 65–77.
, . 1993. Materials management in decentralized supply
chains. Oper. Res. 41(5) 835–847.
, R. So, C. S. Tang. 2000. The value of information in a two-
level supply chain. Management Sci. 46(5) 626–643.
, C. S. Tang. 1997. Modeling the costs and bene?ts of delayed
product differentiation. Management Sci. 43(1) 40–53.
, S. Whang. 1998. Value of postponement from variabil-
ity reduction through uncertainty resolution and forecast
improvement. T. Ho, C. Tang, eds. Product Variety Management:
Research Advances. Kluwer Publishers, Boston, MA, 66–84.
, . 2002. The impact of secondary market on the supply
chain. Management Sci. 48(6) 719–731.
, V. Padmanabhan, S. Whang. 1997. Information distortion in
the supply chain: The bullwhip effect. Management Sci. 43(4)
546–558.
Lundegaard, K. 2001. Bumpy ride: Supply chain management
sounds beautiful in theory; in real life, its a daunting task. Wall
Street J. (May 21).
Miyoaka, J., W. Hausman. 2001. Why use stale forecasts? Because
sometime they perform better! INFORMS Annual Meeting,
Miami, FL.
Nault, B. R., R. K. Tyagi. 2001. Implementable mechanisms to coor-
dinate horizontal alliances. Management Sci. 47(6) 787–799.
Netessine, S., N. Rudi. 2001. Supply chain structures on the
Internet: Marketing operations coordination. Working paper,
Wharton School, University of Pennsylvania, Philadelphia, PA,
http: //omg.simon.rochester.edu /omgHome /Rudi /Serguei /
dropship.pdf.
Palmer, J. W., M. L. Markus. 2000. The performance impacts of
quick response and strategic alignment in specialty retailing.
Inform. Systems Res. 11(3) 241–259.
Peleg, B., H. L. Lee, W. H. Hausman. 2002. Short term
e-procurement strategies versus long term contracts. Production
Oper. Management 11(4) 458–479.
Rajagopalan, S. 2002. Make-to-order or make-to-stock: Model and
application. Management Sci. 48(2) 241–256.
Reinhardt, A. 2001. Tesco bets small and wins big. Bus. Week
(October 1).
Seifert, R., U. Thonemann, S. Rockhold. 2001. Integrating direct and
indirect sales channels under decentralized decision making.
Working paper, IMD, International, Lausanne, Switzerland.
Management Science/Vol. 49, No. 10, October 2003 1405
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
Simchi-Levi, D., D. Wu, Z. Shen. 2003. Supply Chain Analysis in the
eBusiness Era. Kluwer Publishing. Forthcoming.
Sodhi, M. 2001. Applications and opportunities for operations
research in Internet enabled supply chains and electronic
martketplaces. Interfaces 31(2) 56–69.
Swaminathan, J. M. 2001a. Supply chain management. Interna-
tional Encyclopedia of the Social and Behavioral Sciences. Elsevier
Sciences, Oxford, England.
. 2001b. Enabling customization using standardized opera-
tions. California Management Rev. 43(3) 125–135.
, H. L. Lee. 2003. Design for postponement. S. Graves,
T. de Kok, eds. Supply Chain Management—Handbook in
OR/MS, Vol. 11. North-Holland, Amsterdam, The Netherlands,
199–226.
, S. R. Tayur. 1998. Managing broader product lines through
delayed differentiation using vanilla boxes. Management Sci.
44 161–172.
, . 2003. Tactical planning models for supply chain man-
agement. S. Graves, T. de Kok, eds. Supply Chain Management—
Handbook in OR/MS, Vol. 11. North-Holland, Amsterdam, The
Netherlands, 423–454.
, S. F. Smith, N. Sadeh. 1998. Modeling supply chain dynamics:
A multi-agent approach. Decision Sci. 29(3) 607–632.
Tayur, S. R., R. Ganeshan, M. Magazine. 1999. Quantitative Models
in Supply Chain Management. Kluwer Publishers, Boston, MA.
Venkatesh, S., J. M. Swaminathan. 2003. Managing product variety
through postponement: Concept and applications. C. Billing-
ton, T. Harrison, H. L. Lee, J. Neale, eds. The Practice of Supply
Chain Management. Kluwer Publishing, Boston, MA, 139–155.
Wellman, M. P. 1993. A market-oriented programming environ-
ment and its application to distributed multicommodity ?ow
problems. J. Arti?cial Intelligence 1 1–23.
Zipkin, P. 2001. The limits of mass customization. Sloan Management
Rev. 42(3) 81–87.
Accepted by Arthur M. Geoffrion and Ramayya Krishnan; received January 2002. This paper was with the authors 17 months for 5 revisions.
1406 Management Science/Vol. 49, No. 10, October 2003
doc_608253257.pdf
Supply chain management is likely to play an important role in the digital economy. In this paper, we first describe major issues in traditional supply chain management.
Models for Supply Chains in E-Business
Jayashankar M. Swaminathan • Sridhar R. Tayur
Kenan-Flagler Business School, University of North Carolina, Chapel Hill, North Carolina 27599
Graduate School of Industrial Administration, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213
S
upply chain management is likely to play an important role in the digital economy. In
this paper, we ?rst describe major issues in traditional supply chain management. Next,
we focus our attention on the supply chain issues of visibility, supplier relationships, dis-
tribution and pricing, customization, and real-time decision technologies that have risen to
importance with the prevalence of e-business. We present an overview of relevant analytical
research models that have been developed in these areas, discuss their contributions, and
conclude with a discussion on future modeling opportunities in this area.
(Supply Chain Management; Electronic Business; Collaboration; Information Sharing; Decision
Support; Supplier Relations; Procurement; Distribution; Customization; Literature Survey)
1. Introduction
It is estimated that e-commerce in the United States
will grow from $72 billion in 2002 to approximately
$217 billion by 2007, according to a recent Forrester
research report (Johnson et al. 2002). While the pre-
dicted numbers may not be exact, there is no doubt
that the rapid growth and adoption of the Internet
has already had a great impact on all aspects of
business, including customer acquisition, marketing,
human resource management, ?nance, information
systems (IS), and operations. Supply chain manage-
ment that has played an important role in traditional
businesses is likely to be crucial in the digital econ-
omy as well (Geoffrion and Krishnan 2001). Keenan
and Ante (2002) note that during the next ?ve years,
collaboration by supply chain partners over the Inter-
net can potentially save $223 billion with the reduc-
tion in transaction, production, and inventory costs.
This could be one of the largest bene?ts of Internet
technology and many ?rms have already begun to
realize these bene?ts. For example, Microsoft used
a Web collaboration tool to bring the Xbox video
game console to market two months ahead of sched-
ule (Keenan and Ante 2002); Cutler-Hammer has dou-
bled pro?ts and increased productivity by 35% for its
con?gured motor control centers and control panels
by coordinating with customers over the Internet
(Bylinski 2001); and Autoliv reduced the plant inven-
tory by 37% by coordinating orders online with sup-
pliers (Lundegaard 2001). The large potential impact
of the Internet on supply chain management makes
the study of supply chain models in e-business timely
and important.
Supply chain management is a vast topic. We
?rst provide a comprehensive de?nition of supply
chain management and then discuss opportunities
and changes created as a result of Internet usage.
1.1. Supply Chain Management
A supply chain is the set of entities involved in
the design of new products and services, procuring
raw materials, transforming them into semi?nished
and ?nished products, and delivering them to the
end customer. This de?nition, or a modi?ed ver-
sion of it, has been used by several researchers (see
Lee and Billington 1993, Swaminathan et al. 1998,
Swaminathan 2001a, Keskinocak and Tayur 2001).
Supply chain management is the ef?cient manage-
ment of the end-to-end process, which starts with
the design of the product or service and ends with
the time when it has been sold, consumed, and
?nally, discarded by the consumer. This complete pro-
cess includes product design, procurement, planning
0025-1909/03/4910/1387
1526-5501 electronic ISSN
Management Science © 2003 INFORMS
Vol. 49, No. 10, October 2003, pp. 1387–1406
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
and forecasting, production, distribution, ful?llment,
after-sales support, and end-of-life disposal. Supply
chain management issues can be classi?ed into
two broad categories: con?guration (design-oriented)
issues that relate to the basic infrastructure on
which the supply chain executes, and coordination
(execution-oriented) issues that relate to the actual
execution of the supply chain.
Con?guration-level issues include the following
topics.
Procurement and Supplier Decisions. How many
and what kinds of suppliers are necessary? Which
parts should be outsourced and which should be kept
in house? How can procurement practices be stream-
lined and standardized? How should long-term and
short-term contracts be used with suppliers?
Production Decisions. In a global production net-
work, where and how many manufacturing sites
should be operational? How much capacity should be
installed at each of these sites? What kinds of prod-
ucts and services are going to be supported through
the supply chain? How much variety should be pro-
vided to customers? What degree of commonality is
required across the product portfolio?
Distribution Decisions. What kind of distribu-
tion channels should a ?rm have? How many and
where should the distribution and retail outlets be
located? What kinds of transportation modes and
routes should be used? How should a ?rm exploit
risk-pooling opportunities?
Information Support Decisions. Should enterprise
resource planning software be standardized across
the functional units of a ?rm? Should the supply
chain work on standard protocols or on proprietary
standards?
Coordination level issues include the following
topics.
Material Flow Decisions. How much inventory of
different product types should be stored to realize the
expected service levels? Should inventory be carried
in ?nished form or semi?nished form? How often
should inventory be replenished? Should a ?rm make
inventory decisions or is it better to have the vendor
manage the inventory? Should suppliers be required
to deliver goods just in time?
Information Flow Decisions. In what form is
information shared between different entities in the
supply chain: paper, voice via telephone, electronic
data interchange (EDI)? How much collaboration
occurs among the supply chain partners during new
product development?
Cash Flow Decisions. When do suppliers get paid
for their deliveries? What prices should be charged
for products? What kinds of cost reduction efforts
are taken across the supply chain (or expected of
suppliers)? In a global ?rm, in which currency will a
supplier be paid?
It is clear that supply chain management spans sev-
eral functional and geographical areas, introducing
complexities both in terms of design and execution.
Some of the pertinent factors that complicate supply
chain management decisions include the presence of
multiple agents and their sometimes con?icting incen-
tives; uncertainty in demand, supply and produc-
tion distribution process; asymmetry of information
related to product design, inventory, costs, demand
and capacity across the supply chain between the
various parties in the supply chain; and lead time
between the different entities in the supply chain.
All these complexities lead to several types of inef-
?ciencies in the supply chain: poor utilization of
inventory assets (Lee and Billington 1992); distor-
tion of information such as the bullwhip effect (Lee
et al. 1997); high stock-outs and nonresponsive sup-
ply chains (Fisher et al. 1994); poor customer ser-
vice due to customization-responsiveness challenges
(Swaminathan and Tayur 1998); and double marginal-
ization that leads to lower pro?ts for the supply chain
(Cachon 2003).
The science related to supply chain management
traces its history back to the early 1950s when sev-
eral researchers were interested in understanding
the optimal policies related to inventory manage-
ment. In one of the ?rst works, Clark and Scarf
(1960) developed models for managing inventories at
multiple echelons. Numerous researchers have stud-
ied related inventory problems under stochastic and
deterministic environments over the last 50 years.
This stream of research (mostly from a centralized
perspective) is concisely captured in the research
handbook edited by Graves et al. (1993). Over the
1388 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
last decade, researchers have studied several other
issues related to supply chain management that
include decentralized multiagent models to analyze
supply chain coordination problems, models that inte-
grate information availability across the supply chain
with logistics decisions, models for supply contracts
and demand forecasting, and models that integrate
product design with supply chain management. A
collection of prominent research in this area is con-
tained in Tayur et al. (1999) and Graves and de Kok
(2003).
1.2. E-Business and Supply Chain
E-business can be loosely de?ned as a business pro-
cess that uses the Internet or other electronic medium
as a channel to complete business transactions.
As classi?ed by Geoffrion and Krishnan (2001),
e-business consists of three areas: (1) consumer-
oriented activity and (2) business-oriented activity
supported by (3) the e-business technology infras-
tructure. The consumer-oriented activities consist
of business-to-consumer, consumer-to-consumer, and
government-to-consumer activities. The business-
oriented activities comprise business-to-business,
business-to-government, and government-to-business
activities. The technology infrastructure relates to net-
work infrastructure, network applications, decision
technologies, and software tools and applications.
Within this broad de?nition of e-business activities,
we will restrict our attention mainly to consumer-
oriented and business-oriented activities, as well as
decision technologies that are employed for supply
chain management.
The Internet has in?uenced the usage of supply
chain models in three ways. First, the Internet has
facilitated increased use of enterprise resource plan-
ning (ERP) and advanced planning and optimization
solutions (APS). Second, the ability to obtain real-
time information and the access to large computer
systems is enabling ?rms to develop detailed (high-
granularity) supply chain models that can be uti-
lized to make real-time decisions. Last, the Internet
has created opportunities to integrate information
and decision making across different functional units,
thereby creating a need for supply chain models that
go beyond a business unit to study the extended
enterprise. This has elevated the role of supply chain
models from being decision-making enablers for a
single business unit to being enablers for driving cor-
porate strategy. Thus, the Internet has greatly elevated
the role of supply chain modeling and analysis within
a ?rm.
The advent of e-business has also created sev-
eral challenges and opportunities in the supply chain
environment. First and foremost, the Internet has
increased the opportunity for consumers to buy prod-
ucts and services without going to a store. Though
the practice of direct selling through catalogs and
phone was in use earlier by a few ?rms, the Inter-
net has made this form of sales more signi?cant.
In a direct sales environment, the ful?llment process
determines how long customers will wait between
sale and delivery. This has made the back-end ful-
?llment process—which mostly depends on supply
chain management—extremely important. Further, in
the electronic environment, customer expectations
in terms of quick and timely delivery have also
increased. At the same time, the Internet has opened
up opportunities for ?rms to share information and
ef?ciently coordinate their activities with other enti-
ties in the supply chain. This has created several
new avenues in traditional supply chain areas. For
example, in supplier selection and procurement, ?rms
have to decide if they should join private or pub-
lic exchanges or develop highly-integrated supply
partnerships. They need to determine if they should
use auction and bidding for contracts and, if so,
which type would be most bene?cial. In distribu-
tion, they need to decide if the ?rm will offer prod-
ucts through the Internet channel and, if so, how this
method would differ from the traditional channel.
This raises the question of how the synergies would
be realized in terms of inventory, transportation, and
distribution. Similarly, the availability of real-time
information has raised important questions such as
the degree to which information sharing protocol
should be standard or proprietary; the amount and
type of information that should be shared with the
rest of the supply chain partners; and the types of
collaborative processes that may be bene?cial. The
degree of change in issues related to the supply chain
spans a huge spectrum from concepts and issues that
Management Science/Vol. 49, No. 10, October 2003 1389
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
have been marginally affected to a whole set of new
issues that have emerged as a result of e-business.
First, several issues related to supply chain man-
agement have not necessarily changed in principle,
although e-business may have had an impact on some
of their parameters. For example, to maintain given
levels of service, a ?rm still needs buffer inventory
or buffer capacity. This has not changed as a result
of the Internet, although the uncertainty involved in
the decision making may have decreased with the
availability of more information. Similarly, a ?rm still
needs to take into account the interplay between ?xed
and variable costs, while making decisions related to
procurement or setting up additional capacity. With
the prevalence of the Internet, the ?rm might more
easily be able to obtain a lower procurement price or
salvage excess capacity through market mechanisms.
Next are existing supply chain issues that have
become important as a result of e-business. For exam-
ple, leveraging risk-pooling concepts can greatly ben-
e?t Internet channels because products may be stored
at fewer locations as compared to a traditional dis-
tribution channel. Amazon.com can store all inven-
tory for the entire U.S. market in ?ve warehouses
as opposed to several hundred retail outlets (hence,
stocking points) that would be needed for similar cov-
erage in the traditional channel. Similarly, mass cus-
tomization has gained a lot of momentum with the
Internet because ?rms can allow customers to interac-
tively specify customizations of their offerings. It has
become more important for ?rms to understand how
to cope with customization in an effective manner.
Finally, in the last few years, a third category of
issues new to supply chain management has emerged.
One example is linking the dynamic pricing of prod-
ucts to the inventory and capacity decisions. Another
is coordinating Internet and traditional distribution
channels in terms of prices as well as information
and product ?ows. Additionally, the advent of elec-
tronic marketplaces and auctions has opened a whole
new set of issues related to procurement and supplier
relationships.
1.3. Focus of this Paper
In this paper, we focus on the last two types of sup-
ply chain issues previously discussed: (1) those that
have increased in importance with the Internet, and
(2) those that are new issues in the e-business envi-
ronment. This paper is not intended to be exhaustive
in coverage, rather our aim is to concentrate on spe-
ci?c areas related to supply chain and e-business and
highlight papers that, in our opinion, have tackled
important issues. For a more exhaustive coverage, the
reader is referred to the research handbook by Simchi-
Levi et al. (2003). In particular, our focus will be on
papers that utilize analytical modeling and operations
research methodology in supply chain planning and
execution. This is a relatively new area, thus, many
of the research papers discussed in this article are
unpublished.
1.3.1. Procurement and Supplier Management.
The e-business paradigm has created an immense
opportunity for ?rms to consolidate their buying
processes (also called e-procurement). The ?rst suc-
cess stories in the area of e-procurement came from
software ?rms Ariba (http://www.ariba.com) and
Commerce One (http://www.commerceone.com),
which primarily dealt with indirect goods and
services. The initial idea in these systems was that
when procurement is consolidated under a single
roof, the ?rm bene?ts from price reductions (a direct
result of quantity discount). E-procurement systems
enable individual employees to become aware of
all quali?ed suppliers and to complete the procure-
ment process quickly and easily. Subsequently, more
advanced models for procurement have evolved that
use auctions to determine which ?rm wins a pro-
curement contract. Each contract is open to bids in
an electronic auction house and quali?ed suppliers
are allowed to participate. Although auctions have
been used for industrial contracts before, execution
was rather cumbersome and, thus, were used only
for large and important contracts. E-business ?rms
such as FreeMarkets (http://www.freemarkets.com, a
business-to-business, third-party auctions ?rm) have
made it much easier to conduct these auctions and,
as a result, ?rms are beginning to use these auctions
more frequently for procurement. This poses new
research issues such as whether a ?rm should use
such auctions for all its components and suppliers;
what the long-term effect of such auctions on supplier
relationship is; how ?rms can use both traditional
1390 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
and auction mechanisms to hedge against risks in a
more ef?cient manner; and how a third party such as
FreeMarkets can ensure that capacities of suppliers
are taken into account in these auctions so that the
contract is executable.
A related phenomenon in the supplier management
area is the formation of industry-wide supply chain
consortia such as e2open (high tech) and Covisint
(automobile). The motivation behind the formation
of these consortia (also sometimes called market-
places) is their ability to provide liquidity to inventory
and capacity present in the extended supply chain
and make it possible for the supply chain partners
to get involved in long-term collaborative planning
and design efforts. For example, a manufacturer with
excess inventory could salvage it in the marketplace.
Similarly, buyers may have the ability to conduct auc-
tions for industrial parts and procure capacity options
from the supply base. There are a number of new
research issues that have evolved as a result of the
above changes. For example, how could one quan-
tify the bene?ts of joining such a consortium for a
?rm; how many ?rms of the same type are likely to
stay in a consortium; how could the different ?rms
use the liquidity and options in the marketplace to
improve their operational performance. We will dis-
cuss research on the issues of procurement and sup-
plier management in §2.
1.3.2. Visibility and Information Sharing. The
prevalence of ERP allows ?rms to have access to
data across their supply chains, which could be used
for gaining better ef?ciency and effectiveness (Sodhi
2001). The ability to access information from various
parts of the organization has helped ?rms to stream-
line their business processes and reduce inef?cien-
cies. Although ERP systems were implemented before
the boom in e-business, their potential could not be
explored and expanded due to lack of common stan-
dards and cost of access. The growth of e-business
allows and requires that the information made avail-
able from the ERP systems be shared with other ?rms
in the extended supply chain through the Internet.
This enables ?rms to coordinate and collaborate with
their suppliers and customers as well as synchronize
their in-house operations.
The ability to access information across the supply
chain and use it in real time provides various oppor-
tunities. Inventory requirements for buffer stocks are
likely to be lower, because the uncertainty in forecasts
and demand can be reduced across the supply chain.
Allocation of inventory to different retail outlets or
customers as part of order ful?llment can be done
more effectively when there is visibility about the
number and type of inventory located at the different
sites in the supply chain. As more supply chain exe-
cution information becomes available, ?rms can plan
for future operations using advanced planning and
optimization tools. The ability to share information
creates an opportunity for ?rms to have collaborative
planning and design, which removes the inef?cien-
cies in these processes. This has opened up several
new issues for researchers. For example, if forecasts
are shared more often than before, how should they
be used more effectively; how can collaborative fore-
casts be generated and used across the channel; how
can a ?rm use the demand, inventory, or supply infor-
mation from other partners more effectively. We will
discuss research on these issues in §3.
1.3.3. Pricing and Distribution. Physical retail
stores have traditionally been the most common mode
of distribution. Catalog ?rms have traditionally been
considered a rather niche market for only certain types
of products. E-business has opened up a completely
new dimension to distribution by enabling ?rms to
use alternative distribution channels in addition to
brick-and-mortar stores. Also, the ability to share
information across the supply chain has made it eas-
ier for ?rms to try to coordinate the ?ow of materi-
als across multiple channels. In a rush to exploit this
opportunity, several ?rms during the last few years
have set up Internet stores and created new busi-
nesses (such as online grocery). However, many of
them paid little or no attention to logistics associated
with those operations and, as a result, did not go far.
For example, online grocers grossly underestimated
the transportation costs associated with the “last
mile” delivery and the effect of capacity utilization on
economies of scale (Reinhardt 2001). Traditional ?rms
such as IBM, Barnes & Noble, and Wal-Mart also
faced challenging issues during this period when they
created their own Internet operations. These questions
Management Science/Vol. 49, No. 10, October 2003 1391
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
included how the Internet operation should differ
from traditional practice; how the products should be
priced across these two channels; whether all prod-
ucts should be offered on both channels; and what
kind of autonomy should be provided to Internet
operations.
The ability to dynamically change prices in the
marketplace is yet another aspect of distribution that
e-business has greatly impacted. In traditional sup-
ply chain operations, price changes were allowed
but would occur only at regular or planned inter-
vals. However, the Internet enables ?rms to more
dynamically adjust prices with little additional effort.
This poses interesting issues related to how often
prices should be changed and how one can effectively
couple dynamic pricing with production or capacity
decisions.
The Internet has compounded one of the bene?ts
of distribution that relates to risk pooling. Because
products may be stored at fewer physical loca-
tions, the bene?ts of risk pooling will lead to lower
inventory requirements. Having a separate inven-
tory depot for Internet operations provides the abil-
ity to adjust inventory allocations between the retailer
and the inventory depot. This raises several impor-
tant research issues. How much inventory should be
stored at the various locations? How can ?rms cre-
ate incentives so that the brick-and-mortar retailers
can collaborate with each other? We discuss pertinent
research related to distribution and pricing in §4.
1.3.4. Customization and Postponement. The In-
ternet has increased the expectation of customers for
complete customization at a nominal charge. Even
before the advent of e-business, ?rms faced the chal-
lenges related to mass customization and high prod-
uct variety, but this has increased immensely over the
last few years. Part of the reason is the Internet pro-
vides an easy and convenient way for customers to
express their preferences, and sometimes even view
the product as it is customized. Firms are ?nding
that it is important not to completely rely on tradi-
tional manufacturing paradigms of inventory build-
ing (make to stock). They are ?ne tuning their produc-
tion process so that they can store inventory of raw
materials or semi?nished inventory and more rapidly
respond to customer demand after the customer has
ordered the product. We will discuss several mod-
els that capture the trade-offs in these operations and
their effect on supply chain performance in §5.
1.3.5. Enterprise Software and Decision Support
Technologies. Last, the Internet has had a great
impact on development of decision technologies that
use the available data across the supply chain. In the
past, although the models were available, their imple-
mentation in real time was almost infeasible. As a
result, they were mainly used for planning purposes.
However, the Internet has facilitated real-time access
to information across the supply chain, making it now
possible to use decision models during execution. A
related phenomenon is the development of intelligent
supply chain software agents that could take appro-
priate actions in real time, thus, streamlining supply
chain operations. In §6, we discuss some of these deci-
sion technologies in greater detail.
2. Procurement and Supplier
Selection
One of the major effects of the Internet on supply
chain practices is in the area of procurement. Firms
now use the Internet, not only to diversify the supply
base and hedge risk, but also to obtain lower costs
through auctions. This has raised important issues
related to supply chain management. At the strategic
level, ?rms need to decide whether they should have
long-term contracts with a few ?xed suppliers or use
auctions and a dynamic supplier base to reduce their
costs. In particular, ?rms need to understand under
what circumstances is it bene?cial to have (1) long-
term relationships; (2) auction-based, short-term rela-
tionships; or (3) a combination of (1) and (2). Another
important decision is whether a ?rm should have one
supplier or multiple suppliers and how that choice
may depend on repetition in purchase. Elmaghraby
(2000) presents an extensive survey of economics-
oriented models that deals with supplier selection and
sourcing strategies in traditional settings. Our focus
will be on supplier selection and procurement models
under uncertainty in demand or supply that incorpo-
rate the ability of the ?rm to use the Internet to make
the process more ef?cient.
1392 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
Peleg et al. (2002) develop a model that considers
the potential discounts that an exclusive long-term
supplier may be able to offer based on learning and
compare that to the savings that might be gener-
ated through an auction, in a two-period setting with
uncertain demand. The strategic partner guarantees
a price of p per unit in the ?rst period and p(1 ?A)
in the second period, where 0 - A - 1. This model
attributes it to the long-term learning effects of an
exclusive supplier. In an auction-based setting, the
?rm still pays a per unit price p in the ?rst period, but
conducts an auction in the second period. The ?rm
only knows the distribution of the minimum price
(which is assumed to be independent of the quan-
tity bought) that could be obtained as a result of the
auction. In the combined strategy, the ?rm decides
to use the exclusive supplier for the most part (by
having a contract to order a minimum amount in
the second period) and orders any remaining amount
from the auction. Under random demand assump-
tions, Peleg et al. (2002) derive the optimal ordering
quantities for the ?rm. They show that there exists a
A beyond which it is optimal for the buyer to prefer a
strategic long-term partnership, as opposed to using
an auction in the second period. Further, depending
on the distribution of price obtained from the auction,
the combined strategy may be superior or inferior to
a pure auction strategy. Finally, the authors explore
the effect of increasing the supplier base (in the auc-
tion) by modeling the bene?ts obtained (due to poten-
tial lower prices) in the distribution of price from the
auction.
In the automotive supply chain, a tier-1 manu-
facturer such as Ingersoll-Rand often has to reserve
capacity from its supplier, worldwidepm.com, which
is a coalition of small powder metal technology man-
ufacturers. Because the coalition is a powerful entity,
the manufacturer may need to develop contracting
mechanisms such as capacity reservation (which is
similar in spirit to the minimum order assumptions
of Peleg et al. 2002). Motivated by the above setting,
Erhun et al. (2001) study the strategic interaction and
questions of participation (from the suppliers), pro?t
sharing, and impact on end customers. They ana-
lyze a two-period model where the downward slop-
ing demand in the second period could be in one of
two states. This uncertainty is resolved in the second
period where demand becomes deterministic. In con-
trast to standard two-period models, they assume that
demand only occurs in the second period. The ?rst
period is only for capacity reservation and demand
resolution. Under this setting, they demonstrate the
existence of a unique subgame perfect Nash equi-
librium under the various models. They show that
the supplier, manufacturer, and the supply chain as a
whole are aided by this additional negotiation period
where demand is actually not realized, but demand
resolution occurs and capacity is reserved. Under
uncertain demand and limited capacity at the sup-
plier, they show that the additional period is bene-
?cial, but the impact on pro?ts (and its sharing) is
more involved and depends on the level of capacity
available.
As more and more ?rms begin to use auctions for
conducting business transactions, it is important that
models be developed that capture the buyer bidder
interactions and tie them to supply chain decisions
and constraints. Motivated by FreeMarkets, Gallien
and Wein (2000) study the design of smart markets
for industrial procurement. They study a multi-item
procurement auction mechanism for supply environ-
ments with capacity constraints. This enables them to
model rational behavior of suppliers in terms of their
responses when they have limited capacity. In partic-
ular, they consider an auction with n suppliers and
m components, each having a ?xed order size. The
objective of the manufacturing ?rm is to minimize the
total cost of procurement across all the components,
taking into account the capacities of the various sup-
pliers. In each round, suppliers give a price bid for
each of the components. The assumption is that the
price bids decrease as the rounds progress. Further, at
the end of each round, the auctioneer provides each
of the suppliers information about allocations and the
best bids for the next round that would maximize the
supplier’s pro?ts, assuming all other suppliers hold
on to their old bid. The auction stops when the new
set of bids from all the suppliers is identical to the
previous round. Under these conditions, the authors
show that such an auction does converge, and they
provide a bound for the pro?ts of the manufacturer.
Management Science/Vol. 49, No. 10, October 2003 1393
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
Further, for special cases with two suppliers, they pro-
vide insights on the impact of initial bids.
Davenport and Kalagnanam (2001) study auction
strategies for sourcing large quantities of direct mate-
rials over the Internet. They study two auction mech-
anisms, the ?rst of which solicits supply curves as
bids to capture quantity discounts. The second auc-
tion aggregates short-term demand across multiple
manufacturing facilities and allows suppliers to pro-
vide bundled all-or-nothing bids. They incorporate
business rules as side constraints in their integer pro-
gramming formulation and use computational results
to study these mechanisms.
While an auction typically serves as a price-
determination mechanism, Jin and Wu (2001a) show
that the auction could also serve as a coordina-
tion mechanism for the supply chain. They demon-
strate that different forms of auction and market
mechanisms change the nature of supplier competi-
tion, thus, the buyer-supplier interaction. In partic-
ular, they consider a two-supplier–one-buyer system
under four different types of market schemes. They
propose a two-part contract auction where the buyer
announces a price-sensitive order function, while the
suppliers compete in an ascending bid side-payment
auction. Channel coordination can be achieved if the
market intermediary strives to reduce the extent of
information asymmetry, while restricting the buyer’s
pro?t on the side payments. Using the insights from
the two-supplier–one-buyer analysis, they rank mar-
ket schemes by their impact on expected channel ef?-
ciency, expected pro?tability for the buyer, expected
pro?tability for the winning supplier, and expected
commission revenue for the market maker.
Lee and Whang (2002) investigate the impact of a
secondary market where resellers can buy and sell
excess inventories. They develop a two-period model
with a single manufacturer and many resellers. The
resellers order products from the manufacturer in
the ?rst period and are allowed to trade invento-
ries among themselves in the second period. They
derive the optimal decisions for the resellers, along
with the equilibrium market price of the secondary
market. Two types of effects are created due to the
secondary market: (1) a quantity effect (sales by the
manufacturer), and (2) an allocation effect (supply
chain performance). They show that the former is
indeterminate, i.e., the total sales volume for the man-
ufacturer may increase or decrease, depending on the
critical fractile. However, the latter is always posi-
tive, i.e., the secondary market always improves allo-
cation ef?ciency. The sum of the effects is also unclear
in that the welfare of the supply chain may or may
not increase as a result of the secondary market.
Finally, the authors present potential strategies for
the manufacturer to increase sales in the presence of
the secondary market. Dong and Durbin (2001) study
other implications of such surplus markets on supply
chain ef?ciency. Keskinocak and Tayur (2001) describe
examples where matching algorithms developed by
operations researchers have been used in market-
places as decision support for matching demand and
supply. Kleindorfer and Wu (2003) present a detailed
analysis of research related to integrating long-term
and short-term contracting using B2B exchanges.
A related phenomenon in the area of procurement
and supplier relationships has been industry-wide
consortia where multiple buyers and suppliers within
an industry join and conduct business. Better transac-
tional ef?ciency has been highlighted as the key ben-
e?t of consortia such as Covisint (automotive) and
Converge (high tech). The dynamics of these entities
are not well understood and pose several important
questions. For example, one could expect that having
multiple suppliers on the same platform is likely to
reduce prices for the buyer, but it is likely to bene-
?t other buyers in the consortium as well. Thus, it is
not clear if one should join the consortium in the ?rst
place. One of the reasons for Dell deciding not to join
either of the high-tech consortia, Converge or e2open,
could be that they do not want to open up their
supply chain processes to competitors. Granot and
Sosic (2001) study this issue using a stylized model
where there are three ?rms whose products have a
certain degree of substitutability. They explore con-
ditions under which formation of three-member and
two-member alliances is optimal. In particular, they
consider a deterministic linear model for the demand
that depends on the prices charged by the ?rm and its
competitors, and the degree of substitutability across
the products. They assume that when alliances are
formed, the procurement cost (or the price paid by
1394 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
the ?rm) strictly decreases, and that the sum of the
reductions due to a ?rm joining a two-?rm coalition
with each of the other ?rms is greater than the reduc-
tion obtained when all three ?rms are in the alliance.
With these decreases in costs, the authors derive con-
ditions under which joining a three- or two-party
coalition is bene?cial for a ?rm. They also identify
conditions under which a ?rm may prefer to be in a
two-?rm alliance to prevent others from forming an
independent two-?rm alliance even though this may
lead to lower pro?ts. The authors also provide several
other insights into the stability of these alliances.
Jin and Wu (2001b) study the formation of supplier
coalitions in the context of a buyer-centric procure-
ment market. They consider a second-price descend-
ing sealed-bid auction and propose a two-stage
auction mechanism that allows suppliers to form
coalitions with one another. Building on the founda-
tions of core games and bidding rings, they explore
the idea of managed collusion, which provides a
means to enhancing bidder pro?tability. They also
identify basic requirements for a valid coalition mech-
anism, including characteristics such as individual
rationality, welfare compatibility, maintaining compe-
tition, and ?nancial balance. They show that such a
mechanism could be constructed so that the buyer
does not lose the advantage from supplier competi-
tion, and a stable coalition structure could be formed.
They propose a pro?t distribution scheme among
members in the supplier coalition and show that
the proposed scheme provides proper incentives such
that (1) the best strategy for a coalition member is to
comply with the coalition agreement, and (2) bidding
the true cost is the best strategy so long as the bids are
uniformly distributed and the bidder’s cost is above
a certain threshold. They also investigate the stable
coalition structure under the proposed mechanism,
and show that under symmetric information there
exists one unique strongly stable coalition structure.
Butler et al. (1997) suggest that the prevalence
of the Internet will lead to a reduction in costs of
interaction among ?rms that may make horizontal
alliances (i.e., alliances between similar businesses)
more attractive. Nault and Tyagi (2001) examine
coordination mechanisms that achieve an alignment
between the success of the alliance and its individual
members. They consider coordination mechanisms for
a horizontal alliance characterized by the following
features: (1) ?rms in the alliance can exert effort only
in their local markets to increase customer demand
for the alliance, (2) customers are mobile, and a cus-
tomer living in a given alliance member’s local area
may have a need to buy from some other alliance
member, and (3) the coordination rules followed by
the alliance determine which ?rms from a large pool
of potential member ?rms join the alliance, and how
much effort each ?rm joining the alliance exerts in
its local market. In this horizontal alliance setup,
they consider the use of two coordination mecha-
nisms: (1) a linear transfer of fees between mem-
bers if demand from one member’s local customer is
served by another member, and (2) ownership of an
equal share of the alliance pro?ts generated from a
royalty on each member’s sales. The authors derive
conditions on the distribution of demand external-
ity among alliance members to determine when each
coordination mechanism should be used separately,
and when the mechanisms should be used together.
3. Visibility and Information
Sharing
The Internet has made it easier to share informa-
tion among supply chain partners. The current trend
in the industry is to try to leverage the bene?ts
obtained through information sharing (also called vis-
ibility) across the supply chain to improve opera-
tional performance, customer service, and solution
development. The notion of lack of information in the
supply chain and the resulting bullwhip effect was
?rst studied by Lee et al. (1997). Chen et al. (2000)
quantify the effect of forecasting and lead times on
the bullwhip effect under stylized supply chain set-
tings. A number of other papers have dealt with the
?ow of information in the supply chain and the effect
of that on its performance. We will only focus on a
few papers in the interest of space.
Gavirneni et al. (1999) consider the role of infor-
mation in a two-stage capacitated supply chain under
three types of information ?ow: (1) no information is
shared with the supplier (2) the supplier knows the
end-item demand distribution and the retailer uses
Management Science/Vol. 49, No. 10, October 2003 1395
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
an (s, S) policy, and (3) the supplier has complete
information about the inventory state of the retailer.
The authors show the optimality of order-up-to poli-
cies for ?nite and in?nite horizon models. Through a
computational analysis, they study the role of infor-
mation availability and the savings obtained under
different operational conditions. Lee et al. (2000)
study the value of information in a two-level supply
chain with nonstationary end demand and show that
value of information could be high, particularly in
cases where demand may be correlated over time. The
effect of vendor managed inventory (VMI) systems,
where the buyer shares demand information with the
supplier who, in turn, manages the buyer’s inventory,
have also been extensively studied by researchers
(see Cetinkaya and Lee 2000 and Cheung and Lee
2002).
A related issue in supply chain information shar-
ing is sharing forecast information. It is highlighted
in the popular press that one of the reasons for Dell’s
success is its ability to transmit timely and accurate
forecasts to its suppliers. Because forecasts are not
always accurate, ?rms may pass demand information
in the form of bands (lower and upper estimates) that
may get more re?ned as one gets closer to the actual
period of ordering. It is important to understand
how frequent forecasts can be used to improve sup-
ply chain performance. Kaminsky and Swaminathan
(2001) present a model for forecast evolution that cap-
tures two notions related to forecasts: (1) forecasts are
not exact and (2) forecasts over longer horizons are
less certain than those over shorter horizons. With this
forecast evolution model, they develop a capacitated
production planning model for a single product with
terminal demand. After showing that the optimal pro-
duction policy is order up to, they develop heuristics
for the problem and characterize their performance.
Through a detailed computational study, the authors
show that these heuristics get close to the optimal
solution with holding costs (less than 0.5% away from
optimal on average under the conditions studied)
and provide insights on the effect of early, interme-
diate, and late information updates in forecasts on
optimal costs. Although it is reasonable to assume
that forecasts do get better as one gets closer to the
decision epoch in most cases, Cattani and Hausman
(2000) demonstrate situations when forecast updates
may not get better over time. Miyoaka and Hausman
(2001) study a two-stage supply chain where using
old forecasts may be better.
Although sharing forecasts can be bene?cial in most
cases, and the Internet does facilitate the sharing of
such information, con?icting objectives may cause
the forecasts to be distorted. Celikbas et al. (1999)
address this issue in a intra?rm setting where mar-
keting and manufacturing units may have con?ict-
ing objectives regarding forecasts. Marketing, being
a revenue center, may have an incentive to over-
forecast whereas manufacturing, being a cost center,
may have an incentive to underproduce as compared
to the forecast. The authors study a decentralized
setting where demand is uncertain and marketing
provides a forecast to manufacturing, which in turn
produces a quantity based on the forecasted amount
and the knowledge about the demand distribution.
They show that by suitably setting penalties for over-
forecast and underproduction, one can coordinate the
system. Chen (1999) considers a supply chain whose
members are divisions of the same ?rm and are man-
aged by different individuals who have local inven-
tory information. He shows that the owner of the
?rm can manage the divisions as cost centers without
compromising the systemwide performance by using
an incentive-compatible measurement scheme based
on accounting inventory levels. Similar to Celikbas
et al. (1999), Chen (1999) shows that it is important
for the upstream members of the supply chain to
have access to accurate customer demand informa-
tion. More recently, Cachon and Lariviere (2001) study
forecast sharing in a two-stage supply chain between
a manufacturer and a supplier, where the manufac-
turer provides an initial forecast and a contract to
the supplier, who in turn invests to set up capacity
in anticipation of demand. The focus of the study
is on two types of compliances: (1) forced compli-
ance, where the supplier is forced to develop a given
capacity once he accepts a contract from the man-
ufacturer and (2) voluntary compliance, where the
supplier can set the capacity optimally. Cachon and
Lariviere (2001) show that compliance regime plays
an important role in the supply chain performance.
The authors ?nd that it is always in the interest of
1396 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
the manufacturer to truthfully share the forecast with
the supplier, particularly when the demand forecast
is high.
A growing movement in the industry is collabora-
tive planning, forecasting, and replenishment (CPFR),
which tries to reduce the differences that may exist
between the different supply chain players (retailer,
distributor, manufacturer) regarding demand forecast
for a product. In CPFR, the two parties (typically
manufacturer and retailer) jointly collaborate to gen-
erate a forecast and plan for that forecast. Such a
forecast, which uses the information available to all
supply chain players, is likely to make the supply
chain operations more ef?cient because: (1) the fore-
cast is coordinated, and (2) the forecast carries richer
information. Aviv (2001) studies a two-stage two-
player cooperative supply chain. He assumes that the
supply chain is decentralized so that information such
as inventory position and individual forecasts are
only locally available. Under collaborative forecast-
ing, Aviv (2001) assumes that both the manufacturer
and retailer maintain and use a single forecast process
for production replenishment. He compares the ben-
e?ts of collaborative forecasting and local forecasting
in a setting with stationary demand distribution.
4. Distribution and Pricing
Traditional research in the area of distribution has
mainly focused on three issues: (1) the optimal
location of distribution centers and warehouses,
(2) inventory and allocation in a distribution net-
work, and (3) transportation and routing algorithms.
Such models have been studied by researchers in
operations research (OR) for several years. For exam-
ple, Geoffrion and Graves (1974) study the multi-
commodity warehouse location problem for a major
food ?rm; Eppen and Schrage (1981) study the
inventory allocation decisions in a distribution net-
work; and Bramel and Simchi-Levi (1997) present an
overview of logistics research. During these years,
OR researchers have developed effective solution
methodologies for solving large-scale instances of the
production distribution problem that have enabled
?rms to streamline their supply chains. Arntzen et al.
(1995) describe the implementation of such a model
for reengineering at DEC.
The Internet has provided another channel for dis-
tribution of goods. With the prevalence of business-
to-consumer and business-to-business e-commerce,
several ?rms such as Amazon have had to develop
a comprehensive distribution network for their
business. Clearly, the customer base is more dis-
persed and in many cases, customers may buy goods
that need to be delivered elsewhere (such as a gift).
Neglecting the costs and synergies related to distribu-
tion and transportation have proved fatal to several
early movers in this business. In particular, ?rms that
have been most affected by this were online grocers
(like Peapod and Webvan) and home delivery ?rms
(like Kozmo and HomeRuns). This primarily occurred
because these ?rms had to deal with numerous tight
delivery time windows (mostly customer preferred)
in addition to other logistic complexities, which made
it extremely dif?cult to gain ef?ciency through syner-
gies in distribution and transportation.
From a supply chain standpoint, the integration
of traditional and Internet distribution channels is
attractive because it promises pro?t gains, inventory
reduction, and increased customer service. However,
many of the traditional channels are not necessarily
centrally controlled. Seifert et al. (2001) study inven-
tory coordination contracts and their impact on the
direct and indirect channels. Motivated by an actual
business line at Hewlett Packard, they consider a situ-
ation where the ?rm has N traditional identical retail-
ers and one virtual store and random demand. The
assumption is that when a customer is unable to get
the product at the virtual store, then she is willing to
wait. If there is excess inventory at the retailer out-
let at the end of the period, then it is used to satisfy
the demand. Retailers individually make the inven-
tory decisions in a decentralized system, whereas
the manufacturer makes decisions related to inven-
tory at the virtual store. When the channels are not
integrated, then a menu of buy-back contracts can
coordinate the system. In an integrated supply chain
(where excess inventory from the traditional chan-
nel is used to satisfy excess demand from the virtual
store), Seifert et al. (2001) show that when a trans-
fer payment scheme is introduced in addition to a
buy-back scheme, the decentralized system can be
coordinated. Through a detailed computational study,
Management Science/Vol. 49, No. 10, October 2003 1397
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
they provide several managerial insights into pro?ts
and inventory as a function of number of retailers and
degree of coordination.
Anupindi and Bassok (1999) consider a model with
two retailers and one manufacturer, and explore the
bene?ts to the manufacturer and the channel if the
two retailers centralized their inventory in a lost sales
setting. Their study was motivated by some early
pilots by General Motors where the ?rm was explor-
ing how demand and supply availability information
in the distribution channel could be more bene?cially
used. The authors ?nd that centralization of stock
by the retailers need not be in the best interest of
the manufacturer. The true bene?ts to the manufac-
turer may depend on the service levels of the retail-
ers and the degree of substitution (which they term
“search”) of demand between the two retailers. The
authors prove that there exists a threshold level of
search beyond which it is detrimental to manufac-
turer’s pro?ts if the retailers centralize their inven-
tories when the wholesale price is ?xed. For high
search parameters, even the supply chain pro?ts may
decrease as a result of centralization. Even under an
optimal wholesale price setting, they show that a
manufacturer may be better off with a decentralized
system when the degree of search is high.
Netessine and Rudi (2001) study incentive issues
related to drop shipping in business-to-consumer sup-
ply chains. In the drop-shipping method (commonly
used by many e-tailers) the retailer is primarily con-
cerned with customer acquisition (as a result, does not
store any inventory) and the wholesaler takes inven-
tory risk and performs ful?llment. They analyze three
different drop-shipping methods in a multiperiod set-
ting with different power structures: powerful retailer,
powerful wholesaler, and both equally powerful.
These power structures determine the sequence of
actions taken by the ?rms and whether the resulting
equilibrium is Nash or based on Stackelberg. In par-
ticular, they assume that demand is stochastic and the
demand distribution is a linear combination of a ran-
dom component and a function of effort expended
by the retailer for customer acquisition. Netessine
and Rudi (2001) also assume that the mean demand
is increasing and concave in the customer acquisi-
tion spending. They obtain optimal solutions for the
centralized system and traditional (where the retailer
decides both customer acquisition effort and quanti-
ties) and drop-shipping (where the retailer decides on
the effort and the wholesaler decides on the quantity)
methods, and show that these decentralized systems
are not as ef?cient. Further, Netessine and Rudi (2001)
show that traditional contracts based on buy-back
or quantity discounts cannot coordinate the above
decentralized systems. They show that a contract that
also includes a compensation for the retailer’s market-
ing expense can coordinate the system. Finally, several
numerical insights on the problem are provided.
Anupindi et al. (2001) develop a general frame-
work for modeling various issues in decentralized
distribution channels such as drop shipping, ful?ll-
ment houses, and inventory speculators. They con-
sider N retailers who face stochastic demand and hold
stocks locally and/or at one or more central ware-
houses. They assume a given level of substitutability,
so if a retailer is stocked out, a certain portion of
that demand can be satis?ed from existing inven-
tory at other retailers or the central warehouses.
The main decision relates to planning for inventory
and allocation. In a “coopetitive” framework, they
develop the notion of a claim that establishes owner-
ship rights for each unit of inventory in the system
regardless of its location. Thus, by buying a claim on a
unit, a player gets the right to determine how the unit
will be used once demand is realized. This enables
analysis of the inventory decision as a noncooperative
game. Simultaneously, the authors assume that the
shipping and allocation decisions are cooperatively
made. Using the concept of a core (an allocation
scheme, which makes sure that coalitions are stable),
they develop suf?cient conditions for the existence of
one. They also develop conditions for the existence
of a Nash equilibrium for the inventory decisions.
Under the above conditions, they show the existence
of an allocation mechanism that leads to a coordina-
tion across the supply chain and develop conditions
for its uniqueness.
Cattani and Souza (2002) study a model where
an Internet distributor determines the inventory to
stock given that there may be alternative categories of
customers (who may differ in delivery requirements)
and that there may be alternative ways to ship the
1398 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
product. In particular, they assume that there are two
types of customers: (1) those who want air shipment
and (2) those who want ground shipment. Using a
queuing model with Poisson arrivals and exponen-
tial service times, they analyze ?ve different allo-
cation policies. These range from simple ones such
as ?rst-in, ?rst-out (FIFO) to more profound poli-
cies that take into account the inventory position and
prioritize the customer types accordingly. One such
policy serves both types of customers until inven-
tory drops below a threshold S
?
, beyond which only
the air shipments are completed (while the ground
orders may be backlogged or lost). Once the inventory
level increases beyond the threshold (assuming the
machine is always running), the backlogged demand
is upgraded to an air shipment. Using a computa-
tional study, they provide insights into the usefulness
of such ?exibility.
In traditional supply chain research, pricing has
not been an issue because models have been devel-
oped for a single entity whose demand is exogenously
prescribed (of late, there are models for competition
being analyzed; see Cachon 2003). However, with
the advent of the Internet, even a single ?rm needs
to decide how to differentially price products over
the traditional and the Internet channels. Brynjolfsson
and Smith (2000), in their empirical study of book and
CD prices over a 15-month period, concluded that the
Internet prices are 9%–16% lower than prices in con-
ventional outlets. Further, how ?rms price their prod-
ucts on the traditional and the Internet channels may
depend, to a great extent, on the ?rm’s strategy and
the degree of autonomy provided to the Internet arm
of the ?rm. For example, some ?rms see the Internet
as a medium of convenience for customers trying,
instead, to provide useful information on their prod-
ucts, but few discounts. On the other hand, several
?rms (like Wal-Mart) spun off their Internet arms as
independent entities and gave them complete free-
dom to operate their own way.
Cattani et al. (2002b) study optimal pricing for a
?rm that sells on both traditional and Internet chan-
nels, given that the cost and customer preferences in
each segment could differ. In particular, they consider
a linear customer utility model where each customer
has an independent preference for the Internet and
traditional channels. The customer’s choice of channel
depends on the prices of both channels and the cus-
tomer’s individual preference. Under the assump-
tion of uniformly distributed customer preferences,
Cattani et al. (2002b) analyze the impact of vari-
ous pricing and governance issues under monopoly
and competition. Under monopoly, they consider
four different pricing schemes: (1) where traditional
channel prices are not changed and the Internet
channel optimizes its own pro?ts, (2) where tradi-
tional channel prices are unchanged but the Internet
channel prices are set to optimize the joint pro?t, (3)
where both channels price identically but the optimal
price is chosen, and (4) where both prices are jointly
optimized. Cattani et al. (2002b) show that optimal
prices depend, to a great extent, on customer pref-
erences and cost of the two channels. Further, they
establish threshold levels in the difference between
the traditional and Internet costs, beyond which it
is optimal to price one channel higher than the
other. In the competitive case, they consider a tradi-
tional monopolist who is competing against a pure
e-tailer. Cattani et al. (2002b) show the existence of a
unique Nash equilibrium. With a detailed computa-
tional study, the authors provide interesting insights
that, under many conditions, a ?rm, in fact, could
be close to optimal under case (2) above where they
do not change the traditional channel but align the
Internet channel with the existing business. Also, the
joint pricing strategy in case (3) is reasonable when
the costs are not different across the two channels.
Finally, Cattani et al. (2002b) show that case (1), a
strategy adopted by many ?rms over the last cou-
ple of years, could be quite detrimental to the ?rm’s
pro?ts. Using a stylized linear downward sloping
demand function, Huang and Swaminathan (2003)
analytically prove that pro?ts under cases (2) and (3)
are no more than 4% away from the optimal. Further,
they show that under competition, a retailer that has
both the Internet and traditional channels, will price
products higher than a pure Internet retailer.
Druehl and Porteus (2001) examine price compe-
tition between two ?rms, one that has an Internet
channel and another with a brick-and-mortar chan-
nel, under similar demand assumptions as above.
Additionally, they assume that only a fraction of the
Management Science/Vol. 49, No. 10, October 2003 1399
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
population has access to the Internet and is willing to
buy online. They solve for the ?rm’s optimal prices
in Nash and Stackelberg games and relate the com-
petitive outcomes (market structure) in terms of the
degree of innovation. The degree of innovation is
effectively a maximum possible margin (maximum
reservation price minus unit cost) ratio for the two
?rms. As the fraction of consumers willing to buy
online increases, they ?nd that the duopoly outcome
becomes more likely. Thus, even when almost every-
one is willing to buy online, both ?rms are likely to
make positive pro?ts. Using the degree of innova-
tion, they discuss the importance of marketing and
cost control. Balasubramanian (1999) studies a related
problem where he analyzes the competition between
a direct channel and traditional retailers. He models
the customer space in the form of a circle where
the customers are uniformly distributed around the
circumference. The direct channel is located at the
center of the circle equidistant from all the cus-
tomers, thereby having a single location or preference
parameter that determines both the traditional and
direct channel distance for a customer (because the
distance from the center is ?xed at the radius of
the circle). Given this model of the customer space,
Balasubramanian (1999) derives the competitive equi-
librium prices between traditional and direct chan-
nel, and explores the impact of information related to
availability of the product from the direct channel to
customers.
Another related issue is dynamic pricing. With the
advent of the Internet and greater connectivity, ?rms
have a greater ability to dynamically price their prod-
ucts. This approach has been successfully adopted by
?rms in the airline, hotel, and rental car businesses
where the product or service has a perishable nature.
Auctions and bidding are also sometimes referred
to as dynamic pricing models, and have become
prevalent in an e-business environment as discussed
in §3. Another related dynamic pricing model relates
to pricing products based on supply chain charac-
teristics such as of?oading excess capacity. The key
issue is the integration of production and pricing
decisions through the supply chain. Motivated by
an initiative at General Motors, Chan et al. (2001)
study a model that develops dynamic prices, which
are integrated with the production and inventory
control for nonperishable products in a multiperiod
setting. Chan et al. (2001) allow for periodically vary-
ing demand, holding costs, capacities, and produc-
tion costs. In particular, they consider a model where
demand is assumed to be a nonincreasing determin-
istic function of price and the revenue function is
assumed to be concave. Further, they restrict both
production and inventory to be integral values and
have upper and lower bounds on the price the ?rm
can charge in each period. With these assumptions,
they show that the constraint matrix is a polymatroid
and prove that a greedy allocation algorithm for pro-
duction is optimal. Using an extensive computational
study, they offer various interesting insights into
the bene?ts obtained from dynamic pricing under
different operational conditions related to capacity,
demand, and frequency of price changes.
5. Customization and
Postponement
Traditional inventory management models assume
that inventory is stocked and demand is satis?ed
based on availability (called make-to-stock systems).
Swaminathan and Tayur (2003) provide an exten-
sive review of research advances in this area. For
most consumer products, the make-to-stock system
was an appropriate business model. However, the
Internet has resulted in a growing movement toward
selling directly to customers, which has increased
the degree of customization. For manufactured prod-
ucts, ?rms have moved to a build-to-order envi-
ronment or have adopted postponement (also called
delayed differentiation). Venkatesh and Swaminathan
(2003) describe successful implementations related
to postponement across industries and discuss costs
and bene?ts associated with alternative types of
postponement. Swaminathan (2001b) describes exam-
ples that illustrate the increase in customization
due to the Internet. For example, ewatchfactory
(http://www.ewatchfactory.com) enables a customer
to completely customize a watch. This business model
depends on the Internet and a postponement strat-
egy at the ?nal assembly. Another ?rm, Timbuk2
1400 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
(http://www.timbuk2.com), allows consumers to cus-
tomize their computer bags, messenger bags, and lap-
top sleeves. Again, such a business opportunity could
be exploited due to the Internet and a clever combi-
nation of make-to-order and make-to-stock strategies.
In the next few passages, we highlight a handful of
papers that deal with postponement and customiza-
tion models. For a detailed overview on postpone-
ment models, see Swaminathan and Lee (2003).
Lee (1996) describes the most basic version of
the delayed differentiation model where there are
A products and inventory is carried in ?nished form.
All products are customized from the inventory avail-
able at the end of the standard steps of the process.
This model is similar to the one-warehouse multi-
retailer inventory problem studied by Eppen and
Schrage (1981). The basic assumption is that the stan-
dard part of the process takes | time periods to com-
plete, and the remaining T ?| periods correspond to
time for the customization step. This is analogous
to the warehouse lead time of | periods and T ?|
periods of transportation time from the warehouse to
the retailers. All the products are assumed to have
independent and normal demands with mean j
i
per
period and standard deviation u
i
per period. Lee
(1996) analyzes this system with process standard-
ization and addresses the impact of postponement,
which is re?ected in the parameter |. He shows that
the variance of the inventory is decreasing in |. Thus,
postponement will lead to reduction of inventory of
?nished products. Further, the reduction in inven-
tory is greater when the end product demands are
negatively correlated. Lee (1996) also shows that the
reduction in variance is greater when the number
of identical end products is larger. Lee and Whang
(1998) further explore this model by assuming that
demands are not independent and identically dis-
tributed (IID) over time. With non-IID demand, the
value of postponement is more than just the ability to
make product commitments at a later point in time
when realized demands have been revealed. The pro-
gression of demands may also help to improve the
forecast of the future demands.
The above models share an implicit assumption
that the production-distribution process is contin-
uous, and inventory can be stored only in ?n-
ished products form. In general, manufacturing
environments involve a discrete set of operations,
and inventory can be stored immediately following
any one of these stages. Furthermore, the costs asso-
ciated with delaying differentiation have not been
considered in these. Lee and Tang (1997) consider
a model where two products require N sequential
tasks for completion where inventory can be stored
in a buffer after each task, with the buffer after the
Nth task being ?nished goods. The ?rst | tasks are
assumed to have been standardized, i.e., the inven-
tory in the buffer after the |th operation can be used
for customization into either product. The tasks | +1
to N are distinct for the two ?nal products. Thus,
the point of differentiation is right after the |th step.
Under a normal demand assumption for the two
products and a discrete time setting, Lee and Tang
(1997) consider the costs associated with standardiz-
ing any stage of task and develop interesting insights.
Garg and Tang (1997) extend the analysis to a system
with two points of differentiation: the family differen-
tiation, and the product differentiation points.
The above papers assume that the production
distribution process does not have any capacity
constraints. Swaminathan and Tayur (1998) analyze
a ?nal assembly process with production capacity
where inventory is stored in an intermediate form
(vanilla boxes). In addition to the intermediate form,
they allow the two extreme forms of vanilla boxes:
as components, and as ?nished products. Therefore,
this model captures both assemble to order (where
components are stocked and products are assem-
bled from the components after demand is realized)
and make to stock (where inventory is carried in
?nished form only) as special cases. This approach
allows for multiple points of differentiation, as there
is no restriction on the type of vanilla box stored.
The ?nal product demands are assumed to have
any stochastic distribution allowing for correlation
across product demands. Swaminathan and Tayur
(1998) assume that the vanilla box inventory follows
a base stock policy. The inventory is brought up
to that level in every period, and when demand is
realized, products are assembled from vanilla boxes
by adding other components within the production
capacity. Unsatis?ed demand is lost with a penalty
and the remaining inventory of vanilla boxes incurs
Management Science/Vol. 49, No. 10, October 2003 1401
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
holding cost. Using a stochastic integer program and
an ef?cient simulation-based algorithm, Swaminathan
and Tayur (1998) explore the bene?ts of postpone-
ment through vanilla boxes under various settings.
Among other results, they show that postponement
using vanilla boxes outperforms both assemble-to-
order and make-to-stock systems when the assembly
capacity available is neither too slack nor too tight
(representative of most real environments). Further,
they ?nd that the vanilla box approach is extremely
powerful under high variance and negative corre-
lation among product demands. Finally, they pro-
vide examples where stocking two types of vanilla
boxes may be suf?cient for a product family with
10 products, and the performance may be better
than a make-to-stock approach (with all 10 products).
Several other authors since have studied the post-
ponement and the make-to-order versus make-to-
stock problem under various conditions (see Graman
and Magazine 2002, Benjaafar and Gupta 2000).
Rajagopalan (2002) studies a multiitem problem
with setup times between production, limited capac-
ity, and congestion effects. In such an environment,
making an item to order reduces inventory but may
increase the lot size and inventory costs for other
items that are made to stock. Further, lead times
increase due to congestion effects, which may lead to
higher safety stocks for made-to-stock items and poor
service for made-to-order items. Given this tradeoff,
Rajagopalan (2002) develops a nonlinear integer opti-
mization model to determine which products should
be make to order and which should be make to stock
to minimize the total costs, and develops an ef?cient
heuristic for the problem. More recently, Cattani et al.
(2002a), motivated by customization and capacity
issues faced by Timbuk2, develop a two-stage frame-
work involving marketing and operations models to
determine the conditions under which a ?rm bene-
?ts from using ?exible capacity to produce custom
products as demanded each period, and then ?lling
the production schedule with make-to-stock output of
standard products to restock inventory. This mitigates
the “lumpy” demand for custom products, improv-
ing capacity utilization as compared to a focused
approach where standard items are made with ef?-
cient capacity and custom products with ?exible
capacity. Cattani et al. (2002a) explore how capacity
decisions impact demand parameters, and pricing
impacts capacity decisions and, ultimately, production
cost. The authors provide conditions under which
convergence of the two models is achieved, and illus-
trate their framework with data from the messenger
bag manufacturer.
Clearly, enabling customization in the product line
is a great challenge to ?rms. Although the Internet
facilitates a business model around customization,
Zipkin (2001) describes the limits of mass customiza-
tion. Swaminathan (2001b) presents a managerial
framework based on standardizing operations that
could potentially mitigate the negative effects of
customization.
6. Decision Technology
One of the biggest impacts of the Internet on sup-
ply chain practices has been the development of dif-
ferent types of decision technologies, which can be
used in supply chain planning and execution. This
is also a result of the availability of accurate infor-
mation that could increase the usefulness of deci-
sion models to a great extent. i2 (http://www.i2.com)
and Manugistics (http://www.manugistics.com), two
of the leading ?rms in the supply chain planning
software industry, were the ?rst to develop constraint-
based planning approaches (linear and nonlinear pro-
grams) that tapped into data available in ERP systems
to provide better supply chain planning and execu-
tion decisions as compared to earlier manufacturing
resources planning (MRP) systems. With the growth
in e-business, these and several other ?rms such as
Agile, Inc. (http://www.agile.com) have developed
more solutions that leverage the Internet to share
information, and have created a suite of supply chain
solutions. Academics in the supply chain area have
also played an active role in creating newer decision
technologies that use models developed in research,
and have suf?ciently adapted them to be used in
practice (see Table 1).
Clearly, better decision technologies lead to better
and more ef?cient decisions. However, the success of
the implementation of decision technology depends,
to a great extent, on how it is eventually used in
1402 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
Table 1 A Sample of Firms That Have Used Theoretical Advancements
in Logistics and Supply Chain Management in Their Software
Solutions
Company Description
Evant Uses results from stochastic inventory theory
(www.nonstop.com) and daily forecasts to set inventory level
at the retail sites
SmartOps Provides decision support for supply chain
(www.smartops.com) planning (tactical and strategic) and
execution by tapping into real-time data and
using algorithms to determine supply chain
decisions, such as total chain inventory
levels, joint replenishment, and lot sizes
Manhattan Associates Provides a suite of tools that enables real-time
(www.logistics.com) sharing of information about transportation
capacities and helps in reducing inef?ciencies
AspenTech Provides ?rms in the process industry with
(www.aspentech.com) real-time information and analytical and
simulation tools to optimize their most
important processes
4R Systems Provides decision support to forecast and plan
(www.4rsystems.com) for production and inventory management
more ef?ciently in the retail sector
Optiant, Inc. Provides software to help manufacturers
(www.optiant.com) design and con?gure their supply chains
based on inventory and responsiveness
needs
LogicTools Provides optimization-based decision-support
(www.logic-tools.com) tools for logistics and inventory management
MCA Solutions Provides real-time information and decision
(www.mcasolutions.com) support for managing service parts inventory
the business process. Palmer and Markus (2000),
through an empirical study, explore the interrelation-
ship between business process change (in the context
of quick response in the retail industry), degree of
new information technology, and ?rm performance.
Decision technologies based on real-time agents
for supply chain management that use optimization
theory, agent behavior, and economic theory may
become important in the future. Wellman (1993) con-
siders a market-oriented programming approach to
distributed problem solving by deriving the activi-
ties and resource allocations for a set of computa-
tional agents. He considers the multicommodity ?ow
problem and shows that careful construction of the
decision process, according to economic principles,
can lead to ef?cient distributed resource allocation,
and that the behavior of the system can be meaning-
fully analyzed in economic terms. Swaminathan et al.
(1998) propose the concept of intelligent agents in the
supply chain that may have local and global informa-
tion about the world around them and take appro-
priate actions (using decision models), depending on
the situation. They provide a framework of agents
and decision rules. With the Internet, it is possible
to create such software agents that may be able to
tap into required information from other agents and
entities in the supply chain. More recently, Fan et al.
(2003) discuss the design principle of a decentral-
ized supply chain organization that uses the auction
market as the coordinating mechanism. They present
a bundle-trading market that can auction multiple
resources in bundles. They show that such a system
could provide the right incentive for all the partici-
pants to act in the best interest of the supply chain.
This will take supply chain decision technology to
the next level where distributed decision making may
become more prevalent.
7. Concluding Remarks
In this paper, we have provided an overview of
relevant analytical models in the area of e-business
and supply chain management. The topic is broad,
so we restricted our attention to areas in the sup-
ply chain, which have been signi?cantly impacted
by the advent of the Internet. We concentrated our
efforts on a few papers that, in our opinion, have
addressed important issues. Our focus has mainly
been on research in operations management, opera-
tions research, and management science. We recog-
nize that papers in marketing, information systems,
economics, and other areas address other important
issues related to electronic supply chains.
As ?rms strive to adapt their supply chains to
the e-business environment, they need to decide how
much and in what sequence they will be investing
in various supply chain efforts to leverage the ben-
e?ts of the new technology. Traditional models have
focused on ?nding the optimal strategy for a single
parameter (such as inventory, forecast, lead time,
capacity, price). However, there is a growing need
for models that can provide insights into the relative
impact of altering the different parameters in the sup-
ply chain. For example, a ?rm with limited resources
Management Science/Vol. 49, No. 10, October 2003 1403
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
may need to know whether it is better to invest in
vendor-managed inventory or reduce the lead time
with existing suppliers. Similarly, research that helps
?rms identify supply chain changes that have quick
bene?ts versus longer term results will be in demand.
As more and more ?rms begin to integrate their
online and traditional operations and share more
information over the Internet, real-time supply chain
management models on product life-cycle manage-
ment, dynamic pricing and production coordination,
integrated models for supply consortia, and the coor-
dination of Internet and traditional channels, are
going to become all the more signi?cant.
Acknowledgments
The authors thank the editors, an anonymous associate editor, and
four anonymous referees who provided valuable input and com-
ments that have contributed to improving the content and exposi-
tion of this paper.
References
Anupindi, R., Y. Bassok. 1999. Centralization of stocks: Retailers
versus manufacturers. Management Sci. 45(2) 178–191.
, , E. Zemel. 2001. A framework for analysis of decentral-
ized distribution systems with implications for e-commerce.
Manufacturing Service Oper. Management 3(4) 349–368.
Arntzen, B., G. Brown, T. Harrison, L. Trafton. 1995. Global supply
chain management at DEC. Interfaces 25(1) 69–94.
Aviv, Y. 2001. The effect of collaborative forecasting on supply chain
performance. Management Sci. 47(10) 1326–1343.
Balasubramanian, S. 1999. Mail vs. mall: A strategic analysis
of competition between direct marketers and conventional
retailers. Marketing Sci. 17(3) 181–195.
Benjaafar, S., D. Gupta. 2000. Make-to-order, make-to-stock or delay
product differentiation?—A common framework for model-
ing and analysis. Working paper, Department of Mechanical
Engineering, University of Minnesota, Minneapolis, MN.
Bramel, J., D. Simchi-Levi. 1997. The Logic of Logistics. Springer-
Verlag, New York.
Brynjolfsson, E., M. D. Smith. 2000. Frictionless commerce? A com-
parison of Internet and conventional retailers. Management Sci.
46(4) 563–585.
Butler, P., T. W. Hall, A. M. Hana, L. Mendonca, B. Augus-
tine, J. Manylika, A. Sahay. 1997. A revolution in interaction.
McKinsey Quart. 1 5–23.
Bylinski, G. 2001. The e-factory catches on. Fortune (July 23).
Cachon, G. 2003. Supply chain coordination with contracts.
S. Graves, T. de Kok, eds. Supply Chain Management—
Handbook in OR/MS, Vol. 11. North-Holland, Amsterdam, The
Netherlands.
, M. Lariviere. 2001. Contracting to assure supply: How to
share demand forecasts in a supply chain. Management Sci.
47(5) 629–646.
Cattani, K., W. Hausman. 2000. Why are forecast updates often
disappointing? Manufacturing Service Oper. Management 2(2)
119–127.
, G. Souza. 2002. Inventory rationing and shipment ?exi-
bility alternatives for direct market ?rms. Production Oper.
Management 11(4) 441–457.
, E. Dahan, G. Schmidt. 2002a. Spackling: Smoothing make-
to-order production of custom products with make-to-
stock production of standard items. Working paper, Kenan-
Flagler Business School, University of North Carolina, Chapel
Hill, NC.
, W. Gilland, J. M. Swaminathan. 2002b. Coordinated pricing
on traditional and Internet channels. Working paper, Kenan-
Flagler Business School, University of North Carolina, Chapel
Hill, NC.
Celikbas, M., J. G. Shanthikumar, J. M. Swaminathan. 1999. Coor-
dinating production quantities and demand forecasts through
penalty schemes. IIE Trans. 31(3) 851–864.
Cetinkaya, S., C. Y. Lee. 2000. Stock replenishment and shipment
scheduling for vendor managed inventory. Management Sci.
46(2) 217–232.
Chan, L., D. Simchi-Levi, J. Swann. 2001. Dynamic pric-
ing for manufacturing with stochastic demand and dis-
cretionary sales. Working paper, Massachusetts Institute of
Technology, Cambridge, MA.http://www.isye.gatech.edu/
˜jswann/publications/SwannManuDynPricing.pdf.
Chen, F. 1999. Decentralized supply chains subject to information
delays. Management Sci. 45(8) 1076–1090.
, Z. Dresner, J. K. Ryan, D. Simchi-Levi. 2000. Quantifying
the bullwhip effect in a simple supply chain: The impact of
forecasting lead time and information. Management Sci. 46(3)
436–443.
Cheung, K. L., H. L. Lee. 2002. The inventory bene?t of ship-
ment coordination and stock rebalancing in a supply chain.
Management Sci. 48(2) 300–306.
Clark, A., H. Scarf. 1960. Optimal policies for a multi-echelon inven-
tory problem. Management Sci. 6 475–490.
Davenport, A. J., J. R. Kalagnanam. 2001. Price negotiations for
procurement of direct inputs. B. Dietrich, R. Vohra, eds. Math-
ematics of the Internet: Auctions and Markets. Springer-Verlag,
New York.
Dong, L., E. Durbin. 2001. Markets for surplus components with a
strategic supplier. Workingpaper, OlinSchool of Business, Wash-
ington University, St. Louis, MO.http://www.olin.wustl.edu/
faculty/durbin/Papers/DongDurbinSurplusMarketMS.pdf.
Druehl, C., E. Porteus. 2001. Price competition between an Internet
?rm and a bricks and mortar ?rm. Working paper, Graduate
School of Business, Stanford University, Stanford, CA.
Elmaghraby, W. J. 2000. Supply contract competition and sourcing
policies. Manufacturing Service Oper. Management 2(4) 350–371.
1404 Management Science/Vol. 49, No. 10, October 2003
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
Eppen, G., L. Schrage. 1981. Centralized ordering policies in a
multi-warehouse system with lead times and random demand.
L. B. Schwartz, ed. Multi-Level Production Inventory Control
Systems: Theory and Practice. North-Holland, Amsterdam, The
Netherlands, 51–58.
Erhun, F., P. Keskinokak, S. Tayur. 2001. Sequential procurement in
a capacitated supply chain facing uncertain demand. Working
paper, Graduate School of Industrial Administration, Carnegie
Mellon University, Pittsburgh, PA.http://www.gsia.cmu.edu/
andrew/stayur/feryal.ps.
Fan, M., J. Stallaert, A. B. Whinston. 2003. Decentralized mechanism
design for supply chain organizations using an auction market.
Inform. Systems Res. 14(1) 1–22.
Fisher, M. L., J. H. Hammond, W. R. Obermeyer, A. Raman. 1994.
Making supply meet demand in an uncertain world. Harvard
Bus. Rev. 72(3) 83–92.
Gallien, J., L. Wein. 2000. Design and analysis of a smart market
for industrial procurement. Working paper, Massachusetts
Institute of Technology, Cambridge, MA.http://web.mit.edu/
lwein/www/smarket.pdf.
Garg, A., C. S. Tang. 1997. On postponement strategies for product
families with multiple points of differentiation. IIE Trans.
29 641–650.
Gavirneni, S., R. Kapuscinski, S. Tayur. 1999. Value of information
in capacitated supply chains. Management Sci. 45 16–24.
Geoffrion, A., G. W. Graves. 1974. Multicommodity distribution
system design using benders decomposition. Management Sci.
20(5) 822–826.
, R. Krishnan. 2001. Prospects for operations research in the
e-business era. Interfaces 31(2) 6–36.
Graman, G. A., M. J. Magazine. 2002. A numerical analysis of
capacitated postponement. Production Oper. Management 11(3)
340–357.
Granot, D., G. Sosic. 2001. Formation of alliances in Internet-based
supply exchanges. Oper. Res. Forthcoming.
Graves, S. C., A. G. de Kok. 2003. Supply Chain Management—
Handbook of OR/MS. North-Holland, Amsterdam, The
Netherlands. Forthcoming.
, R. Kan, P. H. Zipkin. 1993. Logistics of Production and
Inventory—Handbook of OR/MS. North-Holland, Amsterdam,
The Netherlands.
Huang, W., J. M. Swaminathan. 2003. Pricing in mixed channels:
Analysis and bounds. Working paper, Kenan-Flagler Business
School, University of North Carolina, Chapel Hill, NC.
Jin, M., S. D. Wu. 2001a. Supply chain coordination in
electronic markets: Auction and contracting mecha-
nisms. Working paper, Lehigh University, Lehigh, PA,http://www.lehigh.edu/˜sdw1/ jin1.pdf.
, . 2001b. Supply chain coalitions in e-commerce auc-
tions: Validity requirements and pro?t distribution. Working
paper, Lehigh University, Lehigh, PA,http://www.lehigh.edu/
˜sdw1/jin2.pdf.
Johnson, C., K. Delhagen, A. Dash. 2002. US e-commerce: The next
?ve years. Forrester Report (August 27).
Kaminsky, P., J. M. Swaminathan. 2001. Utilizing forecast band
re?nement for capacitated production planning. Manufacturing
Service Oper. Management 3(1) 68–81.
Keenan, F., S. Ante. 2002. The new teamwork. Bus. Week
(February 18).
Keskinocak, P., S. Tayur. 2001. Quantitative analysis of Internet
enabled supply chains. Interfaces 31(2) 70–89.
Kleindorfer, P., D. J. Wu. 2003. Integrating long-term and short-term
contracting via business-to-business exchanges for capital-
intensive industries. Management Sci. 49(11) Forthcoming.
Lee, H. L. 1996. Effective management of inventory and service
through product and process redesign. Oper. Res. 44 151–159.
, C. Billington. 1992. Managing supply chain inventories: Pit-
falls and opportunities. Sloan Management Rev. 33(3) 65–77.
, . 1993. Materials management in decentralized supply
chains. Oper. Res. 41(5) 835–847.
, R. So, C. S. Tang. 2000. The value of information in a two-
level supply chain. Management Sci. 46(5) 626–643.
, C. S. Tang. 1997. Modeling the costs and bene?ts of delayed
product differentiation. Management Sci. 43(1) 40–53.
, S. Whang. 1998. Value of postponement from variabil-
ity reduction through uncertainty resolution and forecast
improvement. T. Ho, C. Tang, eds. Product Variety Management:
Research Advances. Kluwer Publishers, Boston, MA, 66–84.
, . 2002. The impact of secondary market on the supply
chain. Management Sci. 48(6) 719–731.
, V. Padmanabhan, S. Whang. 1997. Information distortion in
the supply chain: The bullwhip effect. Management Sci. 43(4)
546–558.
Lundegaard, K. 2001. Bumpy ride: Supply chain management
sounds beautiful in theory; in real life, its a daunting task. Wall
Street J. (May 21).
Miyoaka, J., W. Hausman. 2001. Why use stale forecasts? Because
sometime they perform better! INFORMS Annual Meeting,
Miami, FL.
Nault, B. R., R. K. Tyagi. 2001. Implementable mechanisms to coor-
dinate horizontal alliances. Management Sci. 47(6) 787–799.
Netessine, S., N. Rudi. 2001. Supply chain structures on the
Internet: Marketing operations coordination. Working paper,
Wharton School, University of Pennsylvania, Philadelphia, PA,
http: //omg.simon.rochester.edu /omgHome /Rudi /Serguei /
dropship.pdf.
Palmer, J. W., M. L. Markus. 2000. The performance impacts of
quick response and strategic alignment in specialty retailing.
Inform. Systems Res. 11(3) 241–259.
Peleg, B., H. L. Lee, W. H. Hausman. 2002. Short term
e-procurement strategies versus long term contracts. Production
Oper. Management 11(4) 458–479.
Rajagopalan, S. 2002. Make-to-order or make-to-stock: Model and
application. Management Sci. 48(2) 241–256.
Reinhardt, A. 2001. Tesco bets small and wins big. Bus. Week
(October 1).
Seifert, R., U. Thonemann, S. Rockhold. 2001. Integrating direct and
indirect sales channels under decentralized decision making.
Working paper, IMD, International, Lausanne, Switzerland.
Management Science/Vol. 49, No. 10, October 2003 1405
SWAMINATHAN AND TAYUR
Models for Supply Chains in E-Business
Simchi-Levi, D., D. Wu, Z. Shen. 2003. Supply Chain Analysis in the
eBusiness Era. Kluwer Publishing. Forthcoming.
Sodhi, M. 2001. Applications and opportunities for operations
research in Internet enabled supply chains and electronic
martketplaces. Interfaces 31(2) 56–69.
Swaminathan, J. M. 2001a. Supply chain management. Interna-
tional Encyclopedia of the Social and Behavioral Sciences. Elsevier
Sciences, Oxford, England.
. 2001b. Enabling customization using standardized opera-
tions. California Management Rev. 43(3) 125–135.
, H. L. Lee. 2003. Design for postponement. S. Graves,
T. de Kok, eds. Supply Chain Management—Handbook in
OR/MS, Vol. 11. North-Holland, Amsterdam, The Netherlands,
199–226.
, S. R. Tayur. 1998. Managing broader product lines through
delayed differentiation using vanilla boxes. Management Sci.
44 161–172.
, . 2003. Tactical planning models for supply chain man-
agement. S. Graves, T. de Kok, eds. Supply Chain Management—
Handbook in OR/MS, Vol. 11. North-Holland, Amsterdam, The
Netherlands, 423–454.
, S. F. Smith, N. Sadeh. 1998. Modeling supply chain dynamics:
A multi-agent approach. Decision Sci. 29(3) 607–632.
Tayur, S. R., R. Ganeshan, M. Magazine. 1999. Quantitative Models
in Supply Chain Management. Kluwer Publishers, Boston, MA.
Venkatesh, S., J. M. Swaminathan. 2003. Managing product variety
through postponement: Concept and applications. C. Billing-
ton, T. Harrison, H. L. Lee, J. Neale, eds. The Practice of Supply
Chain Management. Kluwer Publishing, Boston, MA, 139–155.
Wellman, M. P. 1993. A market-oriented programming environ-
ment and its application to distributed multicommodity ?ow
problems. J. Arti?cial Intelligence 1 1–23.
Zipkin, P. 2001. The limits of mass customization. Sloan Management
Rev. 42(3) 81–87.
Accepted by Arthur M. Geoffrion and Ramayya Krishnan; received January 2002. This paper was with the authors 17 months for 5 revisions.
1406 Management Science/Vol. 49, No. 10, October 2003
doc_608253257.pdf