Applied Mathematics and Computation 168 (2005) 14191431

www.elsevier.com/locate/amc

Logistics networks: A game theory

application for solving the transshipment
problem
Pedro M. Reyes
Hankamer School of Business, Baylor University, P.O. Box 98006, Waco, TX 76798, USA

Abstract

As competition from emerging economies such as China and India puts pressure on
global supply chains and as new constraints emerge, it presents opportunities for
approaches such as game theory for solving the transshipment problem. In this paper
we use the well-known Shapley value concept from cooperative game theory as an
approach to solve the transshipment problem for maintaining stable conditions in the
logistics network. A numerical example is presented to show the usefulness of this
approach.
2004 Elsevier Inc. All rights reserved.

Keywords: Logistics network; Transshipment problem; Game theory; Shapley value

1. Introduction

As competition from emerging economies such as China and India puts

pressure on global supply chains and as new constraint emerge, it presents
opportunities for new approaches such as the game theory approach for

E-mail address: [email protected]

0096-3003/$ - see front matter 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.amc.2004.10.030
1420 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431

solving the transshipment problem. The International Data Corporation re-

ports that businesses worldwide spend $19 billion annually on information
technology for supply chain management. Laseter and Oliver [12] suggest that
the fundamental principles of SCM, i.e., set supply chain policies strategically,
analyze trade-os holistically, and employ cross-functional support systems;
have not changed despite the expanded scope of SCM that now encompasses
strategic sourcing, supplier involvement in product development, and customer
fulllment processes in addition to the movement of materials.
Recent real world examples of these principles illustrate how organizations
have leveraged these ideas for competitive advantage. Dell Computer Corpora-
tion was able to overcome the strategic constraints of the bullwhip eect of
increasing demand variation and forecast error in the upstream supply chain
faced by the other PC manufacturers (who sold through the reseller channel)
by building directly to the customer order and eliminating the retailer. Procter
and Gamble (P&G) found that variability in its Pampers diaper business was
self-imposed through the supply chains pricing structures, incentives, and
planning and ordering processes of its retailers such as Wal-Mart to P&G
and P&Gs suppliers. It sought to reduce the bull-whip eect by eliminating
price promotions that cause demand variations, synchronizing planning cycles,
sharing forecast and demand information across the supply chain and stream-
lining replenishment through programs such as vendor-managed inventory
[11,3,1,13].
Federated planning (where the focus is on minimizing sub-optimization in
the extended supply chain by collaborating to address the trade-os and pos-
sibly even break constraints across the extended enterprise) and collaborative
planning, forecasting, and replenishment have helped retailers such as Kroger
Company coordinate with suppliers such as Unilever PLC (or Ahold USA Inc.
and P&G) to synchronize promotional plans, eliminate redundant regional
facilities and route shipments to minimize total cost and strategically reshape
their distribution networks.
Process-oriented support systems that link across functions break the func-
tional perspective at both the strategic and tactical levels. As Table A.1 illus-
trates, supply chain management has also shifted business focus over the last
decade from cross-functional integration to cross-enterprise. The new ques-
tions are not about best practices or current practices but about next practices
that may facilitate value creation and innovation.

2. Literature review

Supply chain management (SCM), which is also known as a logistics net-

work [19] has been extensively studied in recent years. The logistical network
consists of facilities and distribution options that perform the functions of pro-
 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431 1421

curement of materials, transformation of these materials into intermediate and

nished products, and the distribution of these nished products to customers.
SCM encompasses the management of all these activities associated with mov-
ing goods from raw materials through to the end user. SCM coordinates and
integrates all of these activities into a seamless process. It embraces and links
all of the partners in the chain. In addition to the key functional areas within
the organization, these partners include vendors, carriers, third-party logistics
companies, and information systems providers. The emphasis on improving
supply chain management has become a major objective of the corporate world
because it represents an opportunity to resolve monumental problems that face
corporations and create a mismatch between supply and demand throughout
their supply chains. Hence, successful SCM is the process of optimizing a com-
panys internal practices, as well as the companys interaction with suppliers
and customers, in order to bring products to market more eciently. The real
challenge for companies, then, is to make the right decision about where they
want to position themselves in cost, functionality and delivery performance
with respect to both their customers requirements and their competitors strat-
egies and gambits.
The supply chain payo can come in many forms. It might reduce transac-
tion costs by eliminating unnecessary steps in moving product to market. It
could enhance customer service through closer coordination among vendors
upstream and carriers, distributors, and customers downstream. Or may be
it increases market share with better customer service or lower costs.
The issues of cost, customer service, and quality have become signicant
objective in the development of logistic network strategies by many rms seek-
ing competitive advantage. Beamon [2] presented an overview and evaluation
of the performance measures used in supply chain models. These performance
measures are summarized as high level of eciency, high level of customer ser-
vice, and the ability to respond to a changing environmentwhere either of
these must coincide with the rms strategic goals. Traditionally, such network
ow problems have been studied extensively for analyzing systems. Beginning
with the work of Ford and Fulkerson [6], a great number of algorithms have
been directed for various versions of ow problems. Urban [18] modeled and
analyzed supply contracts with periodical commitment, in which the order
quantities are xed and stationary, with limited exibility to change the order
quantity at a cost to the buyer. The problem was formulated as a mixed-integer
linear program and as a network ow problem and the solution methodology
provided for the general, stochastic problem with consideration given to spe-
cic demand distributions.
Dynamic ow problems originate in many applications, such as production-
distribution systems, communication systems, and logistic transportation sched-
uling. The research approach on dynamic ow problems can be modeled either
as networks with discrete time steps or as networks with time continuously.
1422 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431

The rst approach uses the time-expanded networks to produce the theoreti-
cally or practically ecient algorithms. The second approach considers net-
works with time-varying capacities and costs, and focuses on proving the
existence of optimal solutions. Fleischer and Tardos [5] related these two ap-
proaches by extending some of the polynomial time algorithms that work in
the discrete-time model to solve the analogous continuous-time dynamic ow
problems. Such problems include nding maximum dynamic ows and dy-
namic transshipments.
Spekman et al. [15] conducted empirical research on the concept of collab-
orative links in supply chain management, where by openly sharing informa-
tion, the supply chain partners can facilitate their abilities to jointly meet the
users needs. Collaboration resources are directly relevant in communicating
with counter parties and suppliers, and forecasting can trigger automated pro-
curement functions. This is important for developing a supply chain sourcing
strategy. However, having information about inventory position for sourcing
is not enough. The most important issue to the logistics decision maker is
how to distribute and transport inventory on time and under stable conditions.
This includes supply chain planning information about resource availability
and delivery requirements so routing eciencies and load/carrier consolida-
tions can be realized [16].
One tactic that can be put to use by an e-business is the operation of trans-
shipment channels in the supply chain. Hong-Minh et al. [7] examined the use
of an emergency transshipment channelarguing that there are times in the
real-world scenarios where emergency transshipment may arise due to rush
orders from customers. However, while these types of transshipments do oc-
cur in the logistics network, it can be argued that these emergencies can lead to
added costs and creates an unstable condition in the logistics network. To con-
trol the activities in the logistics network and to make the best possible use of
the systems resources, stable conditions must exist.
Other issues entail dening inventory policy in support of the strategy
[20,21]. One inventory policy in practice is the use of lateral stock transship-
ments (or simply referred as transshipment) between locations at the same ech-
elon level. This collaboration tactic, as explained by Tagaras [17], can reduce
both the logistic networks total cost and increase customer service levels simul-
taneously. He further points out that this collaboration could be either after
the observed demand (emergency lateral transshipment) or before the realiza-
tion of demand (preventive lateral transshipment). The emergency lateral
transshipment (ELT) allows for inventory to be transferred between stocking
points as replenishment stock within the same echelon rather than a direct
delivery from the prior echelon.
In this paper we use the well-known Shapley value concept from cooperative
game theory as an approach to solve the transshipment problem. We believe
that this ts well with the recent SCM business scenarios found in many indus-
 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431 1423

tries. In particular, the use of information technology (IT) as an enabler for

sharing information in order to improve supply chain performance among all
of the players. This is further stimulated by the devolvement of e-commerce
[22]. Examples of industries that have adopted and incorporated the use of IT
and e-commerce for sharing information include Ecient Consumer Response
(ECR) by the grocery industry [10], Quick Response (QR) by the textile industry
[4], and Wal-Mart (a giant discount retailer).

3. Flow games

A ow is a way of sending objects from one place to another in a network.

The objects that travel or ows through the network are called ow units or
units. For example, ow units can be a commodity, nished goods, or informa-
tion. The network is presented as a graph with a set V whose elements are
called vertices, and a set A of pairs of vertices called edges. The graph is de-
noted G = (V, A). In practice, we specify a ow as a directed graph. The vertices
in a directed graph are commonly called nodes, and the directed edges are often
called arcs. The nodes from which units enter through a network are called
source nodes, and nodes to which the ow units are routed to are called sink
nodes. Source nodes oer supply, which is represented by the number of units
available at the node. Sink nodes usually have demand, which is represented by
the number of units that must be routed to them.
Games, which are derived from a ow situation, are called ow games. A
ow game in simple terms is a way of sending objects from one place to an-
other, but in doing this, the cooperation of players should be used. Kalai
and Zemel [8] dene a ow game associated to a system of vertices V (often
called nodes), which are connected by arcs aij where i 5 j. The double subscript
ij denotes the arc from node i to node j. Each arc has a capacity cij. Further-
more, a ow is a vector X = (xij) in which the component xij represents the ow
units moving from node i to node j. There are two nodes that distinguished
from the others and are called the source (s) and the sink (t), which have al-
ready been previously dened. There is also a nite and non-empty set N as
the player set. The arcs are considered as owned by the players. Moreover, a
coalition owns the arcs of its members. The set of coalitions is denoted C.
An n-person cooperation ow game is a function v from the set of coalitions
to the set of real numbers. For a coalition S 2 C, v(S) is dened as the maxi-
mum ow value for coalition S and through the network of its members if it
operates on its own. Which means that v(S) stands for the maximum ow that
S can sustain using its own portion of the network. The function v just dened
is called the characteristic function of the game.
P For a game (N, v), the corePof v is dened by the set of all n-vectors X satisfying
xi vS for all S  N and xi vN . The constraints imposed on the Core
1424 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431

(N, v) ensure that no coalition would have an incentive to split from the grand
coalition N, and do better on its own. Meaning that an allocation of v(N) belongs
to the Core (N, v) and is stable during the cooperation between the players.

4. Flow games example

Consider the logistics network consisting of one supplier (s), three distribu-
tors (a, b, c), and one retailer (t) presented in Fig. 1, in which one commodity is
passing instantly from one source to a sink without any loss at the other nodes
(a, b, c).
The numbers shown on the arcs are the capacities of these arcs. Suppose
that three persons own the arcs (P1, P2, P3). Precisely P1 (the upper arcs) owns
the arcs (s, a) and (a, t) with the capacities shown. P2 (the middle arcs) owns the
arcs (s, b), (a, b), (c, b), and (b, t) with the capacities shown. P3 (the bottom arcs)
owns the arcs (s, c) and (c, t) with the capacities shown. The values are obtained
on the sub-networks associated with the capacities:
v1 19; v2 37; v3 26;

v12 64; v13 45; v23 73;

v123 100:
Now that the game is known, a good question is how much of the 100 units will
be allowed to be acted by each player. Suppose that the players (owners) would
agree to solve this network problem by using the Shapley Value.

5. Shapley value of the game

For a game (N, v) the Shapley Value is the function SHi : (N, v) ! Rn, which
is given by

Fig. 1. Logistics Network.

 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431 1425

X jSj 1!n jSj!

SHi : N; v vS vS fig
S2SN
n!

for all i 2 N, and V are the source.

From the example in the prior section, the Shapley Value is computed as
v1 19; v2 37; v3 26;

v12 64; v13 45; v23 73;

v123 100:
For i = 1, the coalitions containing 1, are S = {1}, {12}, {13}, and {123}
S f1g; jSj 1 ! 1 1!3 1!=3! 1=3;

S f12g; f13g; jSj 2 ! 2 1!3 2!=3! 1=6;

S f123g; jSj 3 ! 3 1!3 3!=3! 1=3:

Therefore,
SH 1 : N ; v
1=3v1 v0 1=6v12 v2 1=6v13 v3
1=3v123 v23
19=3 1=664 37 1=645 26 1=3100 73
38=6 27=6 19=6 54=6 23:
For i = 2, the coalitions are S = {2},{12},{23}, and {123}, and
SH 2 : N ; v
1=3v2 v0 1=6v12 v1 1=6v23 v3
1=3v123 v13
37=3 1=664 19 1=673 26 1=3100 45
74=6 45=6 47=6 110=6 46:
For i = 3, the coalitions are S = {3},{13},{23}, and {123}, and
SH 3 : N ; v
1=3v3 v0 1=6v13 v3 1=6v23 v2
1=3v123 v12
26=3 1=645 19 1=673 37 1=3100 64
52=6 26=6 36=6 72=6 31:
1426 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431

Fig. 2. Logistic Network with Shapley Value.

Hence the x1 + x2 + x3 = 23 + 46 + 31 = 100 and the Shapley Value for the

cooperative game is obtained, SH:(N, v) = (23, 46, 31). The units, which belong
to the three players, as they would travel through the network, are shown in
Fig. 2, where on each arc the numbers written are the number of units allow-
able to each player.
Clearly, what happens is that the second player accepts four units from the
rst player on (a, b), (b, t) and ve units from the third player on (c, b), (b, t).
Finally, as Kalai and Zemel [8,9] point out, the core of the ow game is
non-empty. We may check that the Shapley Value we computed is in the core.
We have:
SH 1 N ; v 23 > 19 v1;

SH 2 N ; v 46 > 37 v2;

SH 3 N ; v 31 > 26 v3;

SH 1 N ; v SH 2 N ; v 69 > 64 v12;

SH 1 N ; v SH 3 N ; v 54 > 45 v13;

SH 2 N ; v SH 3 N ; v 77 > 73 v23;

SH 1 N ; v SH 2 N ; v SH 3 N ; v 100 v123 vN :
Hence, we veried the appurtenance of the Shapley Value to the core, which
means that this is a stable solution for the cooperation of the three players.

6. Extended application: supply chain intelligence

In the most recent years, e-commerce as been evolving from transactional to

analytic, as well as from tactical to strategic. Supply Chain Intelligence (SCI) is
 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431 1427

a new category of enterprise software that promises to provide insight into stra-
tegic issues for an organization. SCI applications enable strategic decision
making by collecting detailed information from every stage of the product life
cycle, e.g., from material procurement to the manufacturing oor and into the
hands of the end consumer where warranty periods may apply. SCI provides a
broad business intelligence layer that complements enterprise resource plan-
ning (ERP) and supply chain management (SCM). It takes a very broad view
of the extended supply chain utilizing atomic-level data, which is requisite to
understanding total cost and long-term quality ramications. For instance,
the demand/supply chain-wide view of SCI can help an organization relate de-
sign decisions (such as which components from which suppliers are selected)
with failure rates in the eld (which, in turn, generate costs due to warranty
obligations). SCI technologies provide answers to questions such as, Do we
need to add more distribution centers to reduce our total shipping costs
for next years projected sales by $15 million? So, SCI applications may be
run only once or a few times a year, in contrast to transaction systems,
which are used all the time. Questions an enterprise seeks to answer with help
from SCI systemssuch as, where are we adding value to the product?
What makes our customers buy our products and services? And what
alternative sourcing and manufacturing options can we use to reduce our total
cost of goods sold?can have far-reaching consequences for the organization
[14].
In addition, SCI can help explain production outages caused by quality or
compatibility problems with specic materials from specic suppliers. These
outages are well worth understanding for an organization, since increasing pro-
duction yield even a small percentage through SCI can lead to millions of dol-
lars in return, without investing in additional expensive manufacturing oor
equipment in order to increase output. Table A.2 in the appendix dierentiates
SCM from SCI.
Supply chains, just like any other processes, must evolve, in order to reach
the maximum eciency. Enterprise Resource Planning (ERP) was the rst step
that industries implemented, and some are still in the process of implementing
it, in order to search for this eciency. Later, as the searches for eciency in
supply chains kept advancing, step number two was created, and called Supply
Chain Management (SCM). This software complemented ERP, principally in
the areas of demand and productions planning services. However, processes
must evolve, and the search for eciency turns into a search for excellence.
This is how step number three has been created, and called Supply Chain Intel-
ligence (SCI). Just as SCM complemented ERP, SCI complements both these
prior steps, converting it in a much broader and specic software tool for the
industry.
1428 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431

Many manufacturers are now implementing this new software, in the search
for better performance, reduction of costs, and increment of yield. In fact, when
SCI is compared with SCM, we realize that it is more strategic than tactical, and
more analytical than transactional. It not only controls all the costs through e-
ciency, but also searches for revenue growth. It assists in the forecasting method
anticipating any problems that might occur instead of just forecasting materials
and productions, and it keeps and presents historic data instead of just present-
ing the daily information. The Shapley Value, once found, can be used to assign
the dynamic ows for equilibrium. We can argue that this is critical for the
transshipment problem, since this is a tactical method for maintaining stable
conditions in the logistic and supply chain networks.

7. Conclusions and directions for future research

Logistics and supply chain managers in the e-commerce environment face

many changes and challenges in their network environment. Performance mea-
sures such as high levels of eciency, high levels of customer service, and the
ability to respond to a changing environment are driving forces that continue
to challenge these decision makerswhen compared to just a decade ago.
Managing these logistic and supply chain networks involve more than develop-
ing a sourcing strategy. It was presented that the primary issue that the decision
maker is challenged with is how to distribute and transport inventory under
stable conditions. While traditional network ow problems have been exten-
sively studied for analyzing such systems, this paper presented a game theory
application, namely the Shapley Value for the cooperation game to the trans-
shipment network problem. Concluding that once we verify the appurtenance
of the Shapley Value, then we know that we have a stable solution for the
cooperation of the players. With the evolution of B2B strategies and e-com-
merce, we see this as opportunities in supply chain intelligence for integrating
supplier-manufacturing data.
Further research could be carried out using additional players in the net-
work. Since the diusion level of supply chain technologies varies from one
supply chain to another, traditional supply chains and electronic supply chains
co-exist in markets. How do parallel, traditional and electronic supply chains
dier? How can they be managed most eectively? Future research can inves-
tigate the economic and performance implications of dierent supply chain
management choices? Are there new or dierent drivers for supply chains antic-
ipating web services and other collaborative tools?
The transshipment problem is a tactical method for maintaining stable con-
ditions in the logistic networks. In particular, a sensitivity analysis for estab-
lishing exibilities in how the e-commerce partners cooperate in the network
could be useful information to the logistics decision maker.
 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431 1429

Appendix A. Tables A.1 and A.2 are as follows

Table A.1
The shift in SCMs business focus to create strategic advantage [12]
Old questions in SCM New questions in SCM
How do we get the various functional areas How do we coordinate activities across
of our company to work together to supply companies, as well as across internal
product to our immediate customers? functions, to supply product to the market?
* Business focus shift from cross-functional
integration to cross-enterprise
How do we minimize the costs our How do we minimize the costs of matching
company incurs in production and supply and demand while continuing to
distribution of our products? reduce the costs of production and distribution?
* Business focus shift from physical
eciency to market mediation
How do we improve the way we supply How can we get earlier demand information
product in order to match supply and or aect the demand pattern to match supply
demand better, given the demand pattern? and demand?
* Business focus shift from supply focus to
demand focus
How should our company design products How should collaborators design the product,
to minimize product cost (i.e., cost of process and supply chain to minimize costs?
materials, production and distribution)?
* Business focus shift from single-company
product design to collaborative, concurrent
product, process and supply-chain design
How can we reduce our companys What new supply-chain and marketing approach
production and distribution costs? would lead to a breakthrough in customer value?
* Business focus shift from cost reduction
to breakthrough business models
How should we organize our companys How should we organize the supply chain to
operations to serve the mass market serve each customer or segment uniquely and
eciently while oering customized provide a tailored customer experience?
products?
* Business focus shift from mass-market
supply to tailored oerings

Table A.2
The dierences between SCM and SCI
Supply chain management (SCM) Supply chain intelligence (SCI)
Largely about managing the procurement Provides a broad view of an entire supply
and production links of the supply chain chain to reveal full product and component
life cycle
Transactional Analytic
Tactical decision making Strategic decision making
Helps reduce costs through improved Revels opportunities for cost reduction,
operational eciency but also stimulates revenue growth
(continued on next page)
1430 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431

Table A.2 (continued)

Supply chain management (SCM) Supply chain intelligence (SCI)
Usually just the SM applications data Integrates supplier, manufacturing and
(as a vertical stovepipe) product data (horizontal)
Records one state of the data representing Keeps historic record
now
Assists in material and production planning What-if forecasting based on historic data
Quanties cost of same materials Enables an understanding of total cost
Shows todays yield but cannot explain Drills into yield gures to reveal what
inuences on it thus provides no help caused the performance level so it can
for improvements be improved
Simple reporting Collaborative environment with
personalizable monitoring of metrics

References

[1] M.J. Bauer, C.C. Poirier, Computer Sciences Corp., L. Lapide, J. Bermudez, AMR Research,
e-Business: The Strategic Impact on Supply Chain and Logistics, Council of Logistics
Management, Illinois, 2001.
[2] B. Beamon, Measuring supply chain performance, International Journal of Operations and
Production Management 19 (1999) 275292.
[3] D. Elmuti, The perceived impact of supply chain management on organizational eectiveness,
Journal of Supply Chain Management 38 (3) (2002) 4958.
[4] P.-P. Dornier, R. Earnt, M. Fender, P. Kouvelis, Global Operations and Logistics: Text and
Cases, John Wiley and Sons, Inc., New York, NY, 1998.
[5] L. Fleisher, E. Tardos, Ecient continuous-time dynamic network ow algorithms, Opera-
tions Research Letters 23 (1998) 7180.
[6] L.R. Ford, D.R. Fulkerson, Flows in Networks, Princeton Univ. Press, Princeton, NJ, 1962.
[7] S. Hong-Minh, S. Disney, M. Naim, The dynamics of emergency transhipment supply chains,
International Journal of Physical Distribution and Logistics 30 (2000) 788815.
[8] E. Kalai, E. Zemel, Generalized network problem yielding totally balanced games, Operations
Research 30 (1982) 9981008.
[9] E. Kalai, E. Zemel, Totally balanced games and games of ow, Mathematics of Operations
Research 7 (1982) 476478.
[10] Kurt Salmon Associates, Ecient Consumer Response: Enhancing Consumer Value in the
Grocery Industry, 1993.
[11] L.R. Kopczak, M.E. Johnson, The supply chain management eect, MIT Sloan Management
Review 44 (3) (2003) 2734.
[12] T. Laseter, K. Oliver, When Will Supply Chain Management Grow Up? Strategy + Business,
Issue 32, Fall, Booz Allen Hamilton Inc., 2003, pp. 3236.
[13] H.L. Lee, S. Whang, Demand chain excellence: a tale of two retailers, Supply Chain
Management Review 5 (MarchApril) (2001) 4047.
[14] R. Reddy, Supply chain intelligence, Intelligent Enterprise 6 (8) (2003) 4445.
[15] R. Spekman, J. Kamau, N. Myhr, An empirical investigation into supply chain management:
a perspective on partnerships, Chain Management: An International Journal 3 (1998) 5367.
[16] T. Stank, T. Goldsby, A framework for transportation decision making in an integrated
supply chain, Supply Chain Management: An International Journal 5 (2000) 7177.
[17] G. Tagaras, Pooling in multi-location periodic inventory distribution systems, Omega: The
International Journal of Management Science 27 (1999) 3959.
 P.M. Reyes / Appl. Math. Comput. 168 (2005) 14191431 1431

[18] T.L. Urban, Supply contracts with periodic stationary commitment, Production and
Operations Management 9 (4) (2000) 400413.
[19] D. Simchi-Levi, P. Kaminsky, E. Simchi-Levi, Designing and Managing the Supply Chain:
Concepts, Strategies, and Cases, second ed., McGraw-Hill, New York, NY, 2003.
[20] P. Alfredsson, J. Verrijdt, Modeling emergency supply chain exibility in a Two-Echelon
inventory system, Management Science 45 (10) (1999) 14161431.
[21] G. Cachon, M. Lariviere, Capacity choice and allocation: Strategic behavior and supply chain
performance, Management Science 45 (8) (1999) 10911108.
[22] M. Weber, K. Prasad, Factors underlying use of point-of-sale and electronic data interchange
in retailing logistics, Supply Chain Management: An International Journal 7 (5) (2002) 311
317.