## Or.bsnku.net

**This article appeared in a journal published by Elsevier. The attached**

**copy is furnished to the author for internal non-commercial research**

**and education use, including for instruction at the authors institution**

**and sharing with colleagues.**

**Other uses, including reproduction and distribution, or selling or**

**licensing copies, or posting to personal, institutional or third party**

**websites are prohibited.**

**In most cases authors are permitted to post their version of the**

**article (e.g. in Word or Tex form) to their personal website or**

**institutional repository. Authors requiring further information**

**regarding Elsevier's archiving and manuscript policies are**

**encouraged to visit:**

**Author's personal copy**

European Journal of Operational Research 209 (2011) 273–284
Contents lists available at ScienceDirect
European Journal of Operational Research
Production, Manufacturing and Logistics
On a multi-period supply chain system with supplementary order opportunity
Xiang Li a, Yongjian Li b,⇑, Xiaoqiang Cai c
a College of Economic and Social Development, Nankai University, Tianjin 300071, PR Chinab Business School, Nankai University, Tianjin 300071, PR Chinac Department of Systems Engineering & Engineering Management, The Chinese University of Hong Kong, Shatin, N.T., Hong Kong
This paper considers a supplementary supply–order system in a multi-period situation. In each period,
Received 27 December 2008
the buyer ﬁrst places an initial order based on the demand prediction; he has the opportunity to place
Accepted 19 August 2010
a supplementary order with the supplier after the demand of that period is realized. The supplier main-
Available online 24 August 2010
tains an inventory, and decides the quantity to be produced and the quantity to be provided for the sup-plementary order in each time period. We formulate the problem as a multi-period inventory game, and
derive the optimal production and order policies for the supplier and buyer, respectively. The existence
Supply chain management
and uniqueness of Nash equilibrium is proved in the generalized multi-period setting, and the closed-
form Nash equilibrium solution is obtained when the parameters are stationary. Numerical study is per-
Game theorySupplementary order
formed to reveal more managerial insights. We ﬁnd that the supplementary supply–order mechanism, ifdesigned properly, can effectively improve the multi-period supply chain performance.

Ó 2010 Elsevier B.V. All rights reserved.

to a level higher than the normal order quantity, to prepare a sup-plementary supply for the buyer. However, if the supplier produces
Consider a supply chain consisting of a supplier and a buyer, in
an exceeded amount of product, she has to pay the holding cost for
which the buyer faces a stochastic demand of product. The buyer
her own leftover inventory after the buyer's supplementary order.

commits an initial order, referred to as ‘‘normal order" in this pa-
To summarize, the buyer decides his normal order quantity at the
per, before the demand realization, and then has an opportunity
beginning of each period, to balance the tradeoff between the ben-
to place a supplementary order with the supplier after the demand
eﬁt of lower inventory and the loss of shortage, while the supplier
is realized. The stochastic nature of the demand usually causes
decides her supplementary supply quantity, to balance the tradeoff
insufﬁcient/excess inventory, resulting in substantial revenue
between the proﬁtability of the higher wholesale price of the sup-
loss/holding costs for the supply chain members. In a decentralized
plementary order and the risk of holding excess inventory.

system the performance of the supply chain members could be
In what follows, we brieﬂy relate our paper to the literature,
even worse because of their different and usually, conﬂicting
including study on the emergency replenishment policy, on the
objectives they attempt to optimize.

quantity ﬂexibility contract, and on the supply chain with dual
Some studies, e.g., Cachon (2004), Dong and Zhu (2007), have
discussed the supplementary supply–order problem in a single-
Numerous researchers have studied the use of emergency
period situation. In this paper we will consider the problem in a
replenishment to raise service level and to mitigate the risk of de-
multi-period situation, which arises from the supply chain man-
mand randomness, such as Chiang (2003, 2010), Chiang and Gut-
agement of many non-seasonal products, like raw materials, com-
ierrez (1998), and Sethi et al. (2001, 2003), among others. These
papers study the optimal policies of periodic or continuous review
commodities. In each period, the buyer ﬁrst submits a normal or-
inventory system under consideration of two replenishment
der to the supplier based on the forecast of demand, and then
modes: fast and slow, where the fast one is more expensive than
places a supplementary order if the normal order cannot cover
the slow one. Papers such as Fisher and Raman (1996, 2001), Eppen
the realized demand. The supplementary order is submitted after
and Iyer (1997), Brown and Lee (2003), Yan et al. (2003), and Choi
the realization of the demand and charged at a higher wholesale
(2007) have developed models for fashion product in which the
price, if the supplier has available units to satisfy the order. To in-
stochastic demand only occurs once but more than one replenish-
crease proﬁt, the supplier might rationally increase her inventory
ment opportunity can be utilized. These papers study the effect ofadvance ordering/production and demand information updating
where the interaction between supply chain members is not an
Corresponding author. Tel.: +86 22 23505341.

E-mail address: [email protected] (Y. Li).

0377-2217/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved.

doi:10.1016/j.ejor.2010.08.019
**Author's personal copy**

X. Li et al. / European Journal of Operational Research 209 (2011) 273–284
There also has been much interest in exploring the impact of
related to the quantity decision in the channel (the buyer
two or more order chances from a supply chain perspective. A
moves ﬁrst and then the supplier follows). This fundamental
mainstream research is on the quantity ﬂexibility contracts based
difference leads to some different results (like the wholesale
on option. These contracts allow the buyer to adjust his order
price boundary) in our paper.

quantities after the normal order, and an option price is paid tothe supplier for this ﬂexibility. For example, Barnes-Schuster
In the paper, we investigate how the supplementary supply–or-
et al. (2002) explores the contract with option in a two-period sup-
der system works in a generalized multi-period setting by the
ply chain, in which the supplier has two production opportunities
game theoretic approach. We derive optimal normal order policy
and the buyer is allowed to increase delivery quantity at the sec-
for the buyer and optimal supplementary supply policy for the
ond period by utilizing the options. Wang et al. (2006) investigates
supplier, respectively, and prove the uniqueness of Nash equilib-
a similar ﬂexible contract strategy, except that the supplier has
rium. When system parameters are stationary over time, the ana-
only one production opportunity. Van Delft and Vial (2001)
lytic solution of Nash equilibrium is obtained, and the condition
presents a stochastic programming approach to analyze the mul-
under which this supplementary supply–order mechanism takes
ti-periodic supply chain contracts with options. Other relevant re-
effect is given. We ﬁnd that the supplementary supply–order
searches include Bassok and Anupindi (1997, 1999), Tsay (1999),
mechanism is a simple and applicable method to allocate multi-
Deshmukh and Lian (2009), Zhao et al. (2010), etc.

period demand risk between the buyer and the supplier, and to
Moreover, Deng and Yano (2002a) models a two-echelon supply
achieve substantial improvement in the supply chain performance
chain where a second purchase opportunity at an anticipated high-
if properly designed.

er price induces the supplier to bear uncertainty risk. Deng and
The rest of paper is organized as follows. In Section 2 we present
Yano (2002b) studies the impact of a late ordering opportunity
the description and assumptions of the model. In Section 3 we for-
on a push system, assuming that the price of the late order is ran-
mulate and analyze the model. We establish an inventory game
dom. Ozer et al. (2007) considers a supply chain with a demand
and prove the uniqueness of its Nash equilibrium. The analytic
forecast update under a wholesale price contract and a dual pur-
equilibrium solution and some properties are explored in the case
chase contract, through which the manufacturer provides a dis-
of stationary parameters. Section 4 presents a numerical study to
count for orders placed before the forecast updates. They show
provide more insights and Section 5 concludes the paper.

that the supply chain performance is better off in the dual purchasecontract than the one in the wholesale price contract under certaincircumstances.

2. Model description
Cachon (2004), and Dong and Zhu (2007) are the literature that
most closely relate to our work. Both papers explore the single-
2.1. Problem description
period case in which the supplier has only one production oppor-tunity and the buyer has two order opportunities. Cachon (2004)
We consider the following problem. In each time period of a
analyzes scenarios where a buyer can order from a supplier before
multi-period horizon, a buyer orders a product from a supplier
the demand realization (push contract) or after realization (pull
and then sells it in a market. The market demand of period t is a
contract), or make a combination of these two (advance-pur-
stochastic variable D(t) with known distribution information. The
chase-discount contract), which is very similar to the supplemen-
buyer has two order opportunities in each period. He places a ‘‘nor-
tary supply–order system in our paper. The main focus of Cachon
mal" order q(t) at the beginning of each period t, to raise his inven-
(2004) is on investigating the Pareto set with three contracts and
tory position from x(t) to y(t). Before the end of period t, he receives
showing that the combining consideration of push and pull con-
the normal order q(t) and demand D(t) is realized. Then, if he ﬁnds
tracts yields higher supply chain efﬁciency. Dong and Zhu (2007)
that there is a shortage to meet the demand, he can submit a ‘‘sup-
studies push, pull, and advanced-purchase discount contracts in
plementary" order, which would be delivered to him immediately
an uniﬁed two-wholesale-price contract framework and shows
to satisfy the demand of that period if the supplier has available
how the channel decision changes along with the two wholesale
inventory. The supplementary order is more costly than the normal
prices. In the above two papers, only the single-period problem
order for the buyer, and it is not obliged to be fulﬁlled by the sup-
has been discussed.

plier. The demand shortage after the supplementary order, if any,
Different from the above studies, our work makes the following
becomes lost sale. However, if the normal order exceeds the real-
ized demand in that period, the buyer carries the unsold units inhis inventory to the next period to meet future demands and pays
(1) We study the supply chain system in a multi-period setting,
the holding cost correspondingly. Therefore, when making the nor-
derive the optimal production and order policy for both par-
mal order decision, the buyer should balance between the cost for
ties, and explore Nash equilibrium of the multi-period
excess inventory and the cost for demand shortage.

inventory game. We show that the supplementary supply–
On the other hand, the supplier has one production opportunity
order mechanism is beneﬁcial for a multi-period system if
in each period and so needs to keep an inventory to meet the pos-
the system parameters are properly designed;
sible supplementary order of the buyer. At the beginning of period
(2) When the parameters are stationary we characterize the
t, the supplier raises her inventory position from xs(t) to a level
analytic solution of the Nash equilibrium, and provide an
equal or higher than the buyer's normal order q(t). The supplier
might produce to a higher level since she can charge a higher price
which the supplementary supply–order option will be
for the fulﬁlled supplementary order. However, if the supplier pro-
duces excessively, she needs to bear the holding cost for her own
(3) The decision interaction in our supplementary supply–order
inventory after the buyer's supplementary order. Therefore, when
system is different from that under the two-wholesale-price
making the production decision, the supplier should balance be-
contract proposed in the literature. In our model we estab-
tween the proﬁtability of the supplementary supply and the risk
lish a Nash game, in which the supplier decides the supply
of excess inventory.

quantity and the buyer decides the order quantity in a
Note that in our problem, the buyer is to raise, at the beginning
simultaneously interactive way, while in the two-whole-
of each period t, his inventory position to y(t) by the normal order,
sale-price contract the buyer is virtually a Stackelberg leader
given the initial inventory x(t). The normal order quantity q(t) is
**Author's personal copy**

X. Li et al. / European Journal of Operational Research 209 (2011) 273–284
(4) At the same time of period t, the supplier raises the product
Notation used in the model.

quantity to the level q(t) + K(t) by producing q(t) + K(t)
xs(t) units, given the initial inventory is xs(t).2 Production isﬁnished before demand is realized and q(t) units are delivered
Stochastic demand in period t
Selling price in period t
to the buyer. The remaining are to meet the buyer's supple-
Wholesale price in a normal order in period t
mentary order of that period. After the supplementary order
Wholesale price in a supplementary order in period t
is fulﬁlled, the leftover product, if any, is held in the supplier's
Production cost in period t
inventory for the use of next period and incurs holding cost to
The buyer's inventory holding cost for leftovers by the end of
the supplier.

The supplier's inventory holding cost for leftovers by the end of
(5) In period t, the production cost is c(t). The supplier
charges the wholesale price w1(t) for the normal order
Salvage value of the leftover for the buyer by the end of the last
and w2(t) for the supplementary order. The selling price
is r(t). As it is not the objective to investigate pricing deci-
Salvage value for the leftover for the supplier by the end of the lastperiod
sions in this problem, we suppose that the buyer is in a
The buyer's inventory position after the normal order in period t;
competitive marketplace thus r(t) is exogenous. The
decision variable
speciﬁcation of wholesale prices is prior to any order or
The supplementary supply quantity; decision variable
production quantity decision, e.g., they are determined in
The buyer's inventory position at the beginning of period t
advance according to the bargaining power between sup-
The supplier's inventory position at the beginning of period t
The system's inventory position at the beginning of period t, equal
prices are given parameters and satisfy r(t) > w2(t) >
Discount factor, 0 < c 6 1
w1(t) > c(t).

(6) For the buyer, unit leftover product at the end of period
t = 0, 1, . . , T 2 incurs inventory holding cost h(t + 1). Weassume that h(t + 1) + w1(t) > cw1(t + 1) for t = 0, 1, . . , T 2;
hence equal to y(t) x(t). Meanwhile, the supplier is to raise her
otherwise, the buyer may deliberately order more units than
inventory position to q(t) + K(t) at the beginning of each period t,
necessary in period t in order to satisfy the demand of the
given the initial inventory x
next period, because the procurement cost in period t has
s(t). Thus the production quantity of
the supplier in period t is q(t) + K(t) x
an economic advantage over period t + 1. At the end of the
s(t). We refer to K(t) as
the possible supplementary supply quantity in period t, which rep-
last period t = T 1, leftover unit has a terminal value sT, in
resents the maximum amount of units to be provided to the buyer
which h(T) + w1(T 1) > csT. The inequation holds as other-
in addition to the normal order. Therefore, K(t) + q(t) is the maxi-
wise the buyer may deliberately order inﬁnite amount in
mum that could be used to satisfy the demand in period t.

the last period T 1 just to obtain the terminal value rather
We list our notation in Table 1 and present in details the events
than to fulﬁll the demand.

and assumptions in the next subsection.

(7) For the supplier, unit leftover product after supplementary
order at the end of period t = 0, 1, . . , T 2 incurs inventory
2.2. Events and assumptions
c(t) > cc(t + 1) for t = 0, 1, . . , T 2; otherwise, the suppliermay deliberately produce more units than necessary in per-
(1) There is a sequence of time points t = 0, 1, . . , T. Period t is the
iod t in order to satisfy the order of the next period, because
time interval from the point t until just before the point t + 1.

the production cost in period t has an economic advantage
The inventories of the buyer and supplier are both operated
over period t + 1. At the end of the last period t = T 1, unit
under a periodic review over the planning horizon.

leftover product has a terminal value ST, in which
(2) Demand D(t) is a continuous random variable with the dis-
hs(T) + c(T 1) > cST. The inequation holds as otherwise the
tribution function F
supplier may deliberately order inﬁnite amount in the last
D(t) and ﬁnite mean. The demands are
independent of each other. Both Cachon (2004) and Dong
period T 1 just to obtain the terminal value rather than
and Zhu (2007) suppose that the demand has increasing
to fulﬁll the order.

generalized failure rate (IGFR) property, while in our paper
(8) The information of distribution FD(t) and all other parameters
there is no such restrictive assumption on the demand
are shared by the buyer and supplier. Both parties seek to
maximize their own total expected proﬁts over all periods,
(3) At the beginning of period t, the buyer raises the inventory
position to y(t) by submitting a normal order q(t) = y(t) x(t), given the initial inventory is x(t). The normal orderq(t) is delivered to the buyer and D(t) is realized by the
3. Model formulation, analysis, and results
end of period t. After the demand is fulﬁlled, the leftoverproduct, if any, is held in the buyer's inventory for the use
3.1. The buyer's problem
of next period and incurs holding cost to the buyer. If anyshortage occurs for the demand, the buyer submits a supple-
First, we formulate a multi-period inventory model for the
mentary order to the supplier, and if the supplier has no suf-
buyer with the supplementary supply quantities given.

ﬁcient units to cover the shortage, the stockout partbecomes lost sale. Without any loss of generality we assume
2 One might argue that due to the simultaneousness of the decisions, the number of
the stockout penalty is 0.1
units at the supplier after production could be less than the normal order, i.e., K(t)could be negative. In fact, we can easily show that the supplier will optimally makethe product quantity equal or larger than the order. The only modiﬁcation we need isto add an emergency procurement option for the supplier, i.e., the supplier will
1 Note that the stockout penalty can be incorporated into our model without any
activate a more expensive procurement if there is any shortage for the normal order.

difﬁculty. Let b(t) be the penalty for unmet demand of each period t. Then all results
This emergency procurement is, however, never activated as in the equilibrium the
in this paper are still valid except that we need to change ‘‘r(t)" to ‘‘r(t) + b(t)".

production is always sufﬁcient to cover the normal order.

**Author's personal copy**

X. Li et al. / European Journal of Operational Research 209 (2011) 273–284
Based on the above description, if the buyer raises the inventory
(3) V+(t, x) (hence, V(t, x)) is a concave and nonincreasing function
position to y in period t = 0, 1, . . , T 1, his proﬁt in that particular
(4) s*(t) is decreasing in w1(t) while increasing in w2(t); s*(t)
decreasing in K(t).

Pðt; y; KÞ ¼ w1ðtÞðy xðtÞÞ þ rðtÞE minðy; DðtÞÞ þ ðrðtÞ
w2ðtÞÞE min½KðtÞ; ðDðtÞ yÞþ hðt þ 1ÞE½ðy DðtÞÞþ:
Proof. Before the proof of this lemma, we note the following two
The ﬁrst term is the purchasing cost of a normal order and the sec-
facts. Fact 1: if f(y) is a concave function of y, then g(x) = max{f(y):
ond one is the revenue gained by the normal order. The third term
y P x} is a concave and nonincreasing function of x. This fact can be
represents the proﬁt gained by the supplementary order and the
seen directly from the geometry of concave function. Fact 2: if g(x)
last one corresponds to the inventory holding cost by the end of
is convex and f(x) is concave and nonincreasing, then f(g(x)) is con-
cave too. This fact can be proved directly from the deﬁnition of
Deﬁne V(t, x) as the maximized expectation of the buyer's proﬁt
concave function. It is also easy to see from (6) that P+(t, y, K) is con-
over periods t, t + 1, . . , T 1, providing x(t) = x in period t. The ter-
minal value V(T, x) = s
The proof of part (1)–(3) is based on the inductive approach on
Tx is a linear function of the leftover product x
by the end of the last period T 1, in which h(T) + w
t. First, consider the case t = T 1. As P+(T 1, y, K) is concave in y
1(T 1) > csT.

The recursion x(t + 1) = (y(t) D(t))+ holds for t = 0, 1, . . , T 1 due
and V+(T, x) = 0, H(T 1, y, K) is a concave function of y and the
to the lost-sale assumption. Hence, we develop the following dy-
optimality of base-stock policy is proved for t = T 1. According to
namic formulation.

Fact 1, V+(T 1, x) is a concave and nonincreasing function of x.

Therefore, (1)–(3) is true for t = T 1.

VðT; xÞ ¼ sTx;
Second, provided that the lemma is true for t = n + 1, we show
that it is also true for n. Now that V+(n + 1, x) is a concave function
Hðt; y; KÞ ¼ w1ðtÞy þ rðtÞE minðy; DðtÞÞ þ ðrðtÞ
by the induction hypothesis and according to Fact 2, E[V+(n + 1
w2ðtÞÞE min½KðtÞ; ðDðtÞ yÞþ hðt þ 1ÞE½ðy DðtÞÞþ
(y D(t))+] is a concave function of y. Remember that P+(t, y, K) is
þ cE½Vðt þ 1; ðy DðtÞÞþÞ;
concave for all t, thus H(n, y, K) is concave and the optimality ofbase-stock policy is proved for period n. Once again according to
Fact 1, V+(n, x) is a concave and nonincreasing function of x.

1ðtÞx þ maxfHðt; y; K Þ : y P xg:
Therefore, (1)–(3) is true for t = n.

Here, H(t, y, K) represents the expectation of total discounted proﬁt
Finally, on part (4), it is easy to see that @2Hðt;y;KÞ ¼ 1 < 0, thus
over periods t, t + 1, . . , T 1, when the inventory position is raised
H(t, y, Y) is submodular with respect to w
to y in period t. Eqs. (3) and (4) hold for all t = 0, 1,
1(t) and y. Therefore s*(t),
which maximizes H(t, y, K(t)), is decreasing in w
that {K(t), t = 0, 1,
1(t). Similarly we
. . , T 1} are given, then the buyer's problem
have @2Hðt;y;KðtÞÞ ¼ F
is to determine the inventory position y(0)
P x(0) to maximize
ðtÞðy þ K Þ FDðtÞðyÞ P 0 and @2Hðt;y;KðtÞÞ
the expected total proﬁt V(0, x(0)) in period 0, and to determine
2ðtÞÞfDðy þ K ðtÞÞ < 0,
which yield the corresponding results,
the optimal inventory position y(t) P x(t) thereafter in everyperiod t = 1,
Lemma 1 states that the optimal normal order policy is up to le-
. . , T 1 to maximize the proﬁt V(t, x(t)) (this prop-
erty is referred to as Bellman's principle of optimality in dynamic
vel s*(t) for the buyer. It also indicates that this level decreases as
theory, see Puterman (1994) for
the wholesale price for normal order is higher, the wholesale price
for supplementary order is lower, and more supplementary supply
is provided. Now let s+(t) be the smallest value of y that maximizesP+(t, y, K). We call up-to-level s+(t) as a myopic policy for the buyer's
normal order as it optimizes the current proﬁt while ignoring thefuture. We show that the optimal solution s*(t) is bounded by
Pþðt; y; KÞ ¼ ½hðt þ 1Þ þ w1ðtÞ cw1ðt þ 1Þy þ ½hðt þ 1Þ
the myopic solution s+(t).

þ w2ðtÞ cw1ðt þ 1ÞED minðy; DÞþ ½rðtÞ w
Lemma 2. For all t:
2ðtÞED minðy þ K ðtÞ; DÞ;
Vþðt; xÞ ¼ Vðt; xÞ w
(1) s+(t) equals y, which satisﬁes
Then, the following recursion is equivalent to (2)–(4):
½hðt þ 1Þ þ w2ðtÞ cw1ðt þ 1ÞFDðtÞðyÞ þ ½rðtÞ w2ðtÞ
FDðtÞðy þ KðtÞÞ ¼ rðtÞ w1ðtÞ;
VþðT; xÞ ¼ 0;
(2) s*(t) 6 s+(t).

; y; KÞ ¼ Pþðt; y; KÞ þ cE½Vþðt þ 1; ðy DðtÞÞþ;
Vþðt; xÞ ¼ maxfHðt; y; KÞ : y P xg:
Proof. Maximizing P+(t, y, K) and solving the ﬁrst order optimalcondition immediately yield result (1). As V+(t, x) is nonincreasing
Eqs. (9) and (10) hold for t = 0, 1, . . , T 1. And after the deﬁnition of
in x, cE[V+(t + 1, (y
D(t))+] is nonincreasing in y. So H0(t, y, K) 6
1(T), we have h(t + 1) + w1(t) > cw1(t + 1) for all t = 0, 1, . . , T 1,
P+0(t, y, K). In addition, H0(t, s+(t), K) 6 P+0(t, s+, K) = 0 6 H0(t, s*(t), K).

according to the assumption h(t + 1) + w1(t) > cw1(t + 1) for t =
Due to the concavity of H(t, y, K) in y, we instantly obtain
0, 1, . . , T 2 and the terminal value assumption.

s*(t) 6 s+(t).

3.2. The supplier's problem
(1) H(t, y, K) is a concave function of y.

For given normal order base-stock level {y(t): t = 0, 1, . . , T 1},
(2) A base-stock policy is optimal for the buyer's normal order in
the supplier's decision is the supplementary supply quantities
period t. The optimal base-stock level s*(t) is the smallest value
{K(t): t = 0, 1, . . , T 1} to maximize her own total proﬁt over all
of y maximizing H(t, y, K).

periods. Note the production quantity for the supplier is
**Author's personal copy**

X. Li et al. / European Journal of Operational Research 209 (2011) 273–284
s(t) = y(t) + K(t) x(t) xs(t) P 0,
maxfXðtÞ; yðtÞg < SðtÞ;
max{0, xs(t) + x(t) y(t)}.

ðXðtÞ yðtÞÞþ
maxfXðtÞ; yðtÞg P SðtÞ:
Denote Ps(t, y, K) as the supplier's expected proﬁt in period t
with the decision variable K(t). We have
(2) S*(t) is increasing in w2(t), but irrelevant with w1(t); S*(t) is
increasing in y(t).

Psðt; y; KÞ ¼ w1ðtÞ½yðtÞ xðtÞ cðtÞ½yðtÞ þ KðtÞ xðtÞ xsðtÞ
þ w2ðtÞE min½KðtÞ; ðDðtÞ yðtÞÞþ
sðt þ 1ÞE½K ðtÞ ðDðtÞ yðtÞÞþþ;
where the ﬁrst two terms represent the revenues of the normal
(1) Rearranging the terms of (17) yields
supply and supplementary supply, respectively; the third term isthe production cost; and the last term is the inventory holding
Pþðt; y; KÞ ¼ ½w
1ðtÞ cðtÞ½yðtÞ xðtÞ þ ½w2ðtÞ
vðtÞE½yðtÞ DðtÞþ ½w2ðtÞ cðtÞyðtÞ
Deﬁne Vs(t, xs) as the maximized expectation of the supplier's
þ ½w2ðtÞ cðtÞ½KðtÞ þ yðtÞ
proﬁt over periods t, t + 1, . . , T 1, providing xs(t) = xs in period t.

The terminal value V
½w2ðtÞ vðtÞE½KðtÞ þ yðtÞ DðtÞþ:
s(T, xs) = STxs is a linear function of the leftover
product x by the end of the last period T 1, in which
In the following we use variable substitution to solve the problem.

hs(T) + c(T 1) > cST. The recursion xs(t + 1) = [K(t) (y(t) D(t))+]+
Denote Y(t) = K(t) + y(t) as the new decision variable representing
holds for t = 0, 1, . . , T 1. Hence, we develop the following dy-
the system inventory position raised to in period t, and xe = xs + x(t)
namic formulation.

as the new state variable representing the initial system inventory.

The dynamic programming (19)–(21) is equivalent to
sðT ; xsÞ ¼ ST xs;
UsðT; xeÞ ¼ 0;
Hsðt; y; KÞ ¼ w1ðtÞ½yðtÞ xðtÞ cðtÞ½yðtÞ þ K xðtÞ
þ w2ðtÞE min½K; ðDðtÞ yðtÞÞþ hsðt þ 1ÞEðK
Gsðt; y; YÞ ¼ Mðt; YÞ þ cE½Usðt þ 1; ðY DðtÞÞþ ðyðtÞ
ðyðtÞ DðtÞÞþÞþ þ cE½Vsðt þ 1; ðK ðDðtÞ yðtÞÞþÞþ;
Usðt; xeÞ ¼ maxfGsðt; y; YÞ : Y P max½yðtÞ; xeg:
Vsðt; xsÞ ¼ cðtÞxs þ maxfHsðt; y; KÞ : K P max½0; xs þ xðtÞ yðtÞg:
Mðt; YÞ ¼ ½w1ðtÞ cðtÞ½yðtÞ xðtÞ
Here, Hs(t, y, K) represents the expectation of total discounted proﬁtover periods t, t + 1, . . , T 1, when the supplementary supply is K
þ ½w2ðtÞ vðtÞE½ðyðtÞ DðtÞÞþ
in period t. Eqs. (13) and (14) hold for all t = 0, 1, . . , T 1. The sup-
½w2ðtÞ cðtÞyðtÞ þ ½w2ðtÞ cðtÞY
plier's problem is to determine the supplementary supply quantity
½w2ðtÞ vðtÞE½ðY DðtÞÞþ;
K(t) P max[0, xs(t) + x(t) y(t)], t = 0, 1, . . , T 1 sequentially, for gi-ven inventory states of the buyer and supplier and the buyer's deci-
is a concave function of Y. Now we declare for all t = 0, 1, . . , T 1,
sion {y(t), t = 0, 1, . . , T 1}.

(a) Gs(t, y, Y) is a concave function of Y.

(b) The optimal decision Y*(t) is characterized by
maxfXðtÞ; yðtÞg < SðtÞ;
vðtÞ ¼ ccðt þ 1Þ h
maxfXðtÞ; yðtÞg maxfXðtÞ; yðtÞg P SðtÞ;
Pþðt; y; KÞ ¼ w
1ðtÞ½yðtÞ xðtÞ cðtÞ½yðtÞ þ K ðtÞ xðtÞ
where S*(t) is the smallest value of Y maximizing Gs(t, y, Y).

2ðtÞE min½K ðtÞ; ðDðtÞ yðtÞÞþ
s(t, xs) is a concave and nonincreasing function of xs.

Note now y(t), t = 0, 1, . . , T 1 are regarded as constants, the
þ vðtÞE½KðtÞ ðDðtÞ yðtÞÞþþ;
proof of the above fact is very similar with Lemma 1 thus omitted.

According to (28), it is now easy to see the optimal decision K*(t)
sÞ ¼ V sðt; xsÞ cðtÞxs:
can be expressed as (22).

Then, the following recursion is equivalent to (12)–(14):
(2) It is easy to see that @2Gsðt;y;YÞ ¼ 1 F
DðtÞðY Þ P 0, so Gs(t, y, Y) is
supermodular with respect to w2(t) and Y. So S*(t), which
maximizes Gs(t, y, Y), is increasing in w2(t). The irrelevance
Hsðt; y; KÞ ¼ Pþðt; y; KÞ þ cE½Vþðt þ 1; ½K ðDðtÞ yðtÞÞþþ;
1(t) can be shown as w1(t) only emerges in the terms
which are regarded to be constant value in (23). We also
sÞ ¼ maxfHsðt; y; K Þ : K P max½0; xs þ xðtÞ yðtÞg:
have @2Gsðt;yðtÞ;YÞ ¼ @2Usðt;yðtÞ;YÞ þ @2Mðt;YÞ P 0. Therefore, S*(t) is
Eqs. (20) and (21) hold for t = 0, 1, . . , T 1. And after the deﬁnition
increasing in y(t).

of c(T) and v(t), we have v(t) < c(t) for all t = 0, 1, . . , T 1, accordingto the assumption hs(t + 1) + c(t) > cc(t + 1) for t = 0, 1, . . , T 2 and
Lemma 3 shows that the supplier's production follows a ‘‘base-
the terminal value assumption.

system-stock" S*(t) policy. That is, if the system inventory position
Denote X(t) = x(t) + xs(t), which represents the initial inventory
is higher than S*(t) after the normal order at the beginning of per-
of the supply chain system at the beginning of period t. We have:
iod t, then the supplier will use the leftover inventory to meet thesupplementary order if necessary. Otherwise the supplier will ac-
tively produce an extra amount to be the supplementary supply,raising the system inventory position to S*(t). In other words, the
(1) There exist constants S*(t), t = 0, 1, . . , T 1, such that the opti-
supplier guarantees the system inventory position to be not less
mal decision K*(t) for the supplier satisﬁes
than S*(t) in period t.

**Author's personal copy**

X. Li et al. / European Journal of Operational Research 209 (2011) 273–284
It is also interesting to note that the benchmark S*(t) is increas-
that. In contrast, in the system without the supplementary supply–
ing in w2(t) and has no relation with w1(t), implying that the sup-
order, the system inventory level is raised up to the classic base-
plementary wholesale price w2(t) is a crucial factor to motivate a
stock level by regarding {w1(t), t = 0, 1, . . , T 1} as the purchase
high service level of the supply chain. This result is due to the fact
costs. The difference between the two sets of levels designates
that w2(t) is the wholesale price for the supplementary order and
the effect of supplementary supply–order mechanism for the sup-
w2(t) > w1(t). In fact, S*(t) indicates the system inventory position
ply chain system.

to be raised when the supplementary supply–order takes effect,
In the equilibrium, we consider the optimal dynamic policy for
and in this situation it is intuitively reasonable that the price of
each player while regarding the decision of his contender as a mul-
supplementary order w2(t) would play a core role from the sys-
ti-dimensional vector, i.e., the player is ignorant about the evolu-
tem's perspective. Furthermore, S*(t) increases as the normal order
tion of the decision process of his contender. Exploring the
up-to-level y(t) is raised. This result, combining with (4) in Lemma
closed-loop equilibrium of this model is an interesting problem
1, shows the sensitivity of one player's decision on the other's, and
for the future study.

forms the basis of the game analysis.

Now let S+(t) be the smallest value of Y that maximizes M(t, Y) in
3.4. The stationary parameter case: The stationary Nash equilibrium
(27). Thus the base-system-stock S+(t) is a myopic policy for the
supplier since it optimizes the current proﬁt while ignoring thefuture.

Suppose the parameters of the problem (all the costs, prices,
and demands) are stationary across the entire planning horizon.

Lemma 4. For all t,
In other words, r(t) = r, c(t) = c, h(t) = h, w1(t) = w1, etc., fort = 0, 1, . . , T, and all D(t),t = 0, 1, . . , T 1 share the same distribu-
(1) SþðtÞ ¼ F1 ðw2ðtÞcðtÞÞ,
tion. Let D denote the generic demand. The boundary condition
DðtÞ w2ðtÞvðtÞ
for the last period is V(T, x) = w
1x for the buyer and Vs(T, x) = cx for
the supplier. This indicates that at the end of last period T 1,
6 y(t) for all t = 0, 1, . . , T 1, the myopic policy is opti-
mal for the supplier, i.e., for all t, the optimal decision K*(t)
the supplier refunds to the buyer at unit price w1 for the buyer's
for the supplier satisﬁes
leftover inventory, and salvage them with her own inventory at
unit price c.

SþðtÞ yðtÞ yðtÞ < SþðtÞ;
yðtÞ P SþðtÞ:
(1) Suppose that the supplier's decision is stationary over all peri-
Proof. Maximizing M(t, Y) and solving the ﬁrst order optimal con-
ods as K(t) = K. The myopic base-stock s+ policy is optimal for
dition immediately yield result (1). S*(t) 6 S+(t) can be proved in
the buyer's order problem, where s+ is characterized by [h +
the similar way with result (2) in Lemma 2. Now we prove result
w2 cw1]FD(s+) + [r w2]FD(s+ + K) = r w1. In other words,
(3). If X(t) 6 y(t), Eq. (26) becomes U
the optimal normal order raises the buyer's inventory position
s(t, xe) = max{Gs(t, y, Y):Y P y(t)}
thus the lower bound constraint of decision Y(t) becomes a con-
stant y(t) and no longer relevant to decision Y(t 1). In this case,
the dynamic programming has a separative property and can be
split to the summation of T disconnected items M(t, Y),t = 0, 1, . . ,T 1, each of which can be optimized independently. Maximizing
(2) Suppose that the buyer's decision is stationary over all periods
each M(t, Y) leads to the optimal policy which is base-system-stock
as y(t) = y. The myopic base-system-stock Sþ ¼ F1
icy is optimal for the supplier's production problem. In otherwords, the optimal supplementary supply quantity is
3.3. The supplementary supply–order system: An inventory game
( Sþ y y < Sþ;
In the above subsections, Lemma 3 characterizes the supplier's
optimal decisions {K(t), t = 0, 1, . . , T 1} given the buyer's deci-sions {y(t), t = 0, 1, . . , T 1}. On the other hand, Lemma 1 indicatesthat the buyer can determine his optimal policy {y(t), t =
Proof. The optimality of myopic policy has been well studied in
0, 1, . . , T 1} for speciﬁed {K(t), t = 0, 1, . . , T 1}. Assume that
the literature on stochastic inventory models when the parame-
the buyer and the supplier are able to commit to their ordering
ters are stationary and the terminal salvage value is linear with
and production decisions, respectively. Therefore we establish a
the slope equals to the procurement cost. For example, see
supply–order game between them, in which each one chooses
Zipkin (2000). The proof of our results is very similar and thus
his/her {y(t)}, {K(t)}, respectively, to maximize his/her own objec-
tive, given the decision of the other one. We call this a y K inven-
Lemma 5 states that in the stationary y K inventory game,
each player's optimal strategy is myopic, and also stationary, ifthe decision of the other player is stationary. Result (2) provides
Theorem 1. For any cost and price parameters r(t), w1(t), w2(t), h(t),
the analytic solution for the optimal system inventory position to
r(t) > w2(t) > w1(t) >
be raised in this stationary parameter case, which is consistent
c(t) > v(t) = cc(t + 1) hs(t + 1),
h(t + 1) + w2(t) > cw1(t + 1),
with Lemma 3 in the generalized case. An interpretation for this re-
1, . . , T 1, there exists an unique Nash Equilibrium {y*(t), K*(t), t =
sult is that in this case the supplementary supply–order mecha-
0, 1, . . , T 1} for the y K inventory game.

nism provides the supplier an opportunity to determine the
Theorem 1 shows the uniqueness of Nash equilibrium in the
inventory position for the entire supply chain by taking on the
multi-period y K inventory game. The proof can be found in the
inventory risk at unit overage cost c v and unit underage cost
supplementary material online. It can be observed that in the sup-
SðtÞ ¼ Sþ ¼
plementary supply–order system, the system inventory position at
is obtained as the threshold for the system
period t is raised up to S*(t) by the supplier whenever it falls below
inventory position.

**Author's personal copy**

X. Li et al. / European Journal of Operational Research 209 (2011) 273–284
Now deﬁne GðwÞ ¼ chþcrvrðccvÞw in [c, r].

ðhþhsÞ2þ4ð1cÞðhcþð1cÞcrþhsrÞ 2 ðc; rÞ solving G(w) = w.

GðwÞ ¼ w () ð1 cÞw2 þ ðh þ cc vÞw ðhc þ cr
vrÞ ¼ 0. Denote H(w) = (1 c)w2 + (h + c c v)w (hc + cr vr).

According to the deﬁnition of v(t) and the stationary assumption,
we have v = cc hs < cc. Also c 6 1, so H(w) is an increasing func-
*G*(*w *)
tion in w. It is easy to verify that H(c) = (c v)(c r) < 0 andH(r) = (r c)[h + (1 c)r] > 0, thus we immediately have thatH(w) = 0 has a unique solution in (c, r). Solving H(w) = 0 yields
w ¼ ðhþhsÞþ ðhþhsÞ2þ4ð1cÞðhcþð1cÞcrþhsrÞ.

In the following we suppose x
Region 1 and Region 2: Supplementary Supply-Order Is Activated
0(0) = x(0) = 0, i.e., the initial
inventory at the beginning of the planning horizon is 0 for both
Region 3: Supplementary Supply-Order Is Not Activated
the supplier and buyer. The following result is obtained immedi-
ately from Lemma 5.

1, w2} map under supplementary-order system.

supply and the problem becomes a multi-period ‘‘selling-to-the-newsvendor" problem under single wholesale price w1. Region 3in Fig. 1 represents this case.

w and w2 > w1, or w1 6
w and w2 > G(w1), then
We note that Dong and Zhu (2007) state that in their single-per-
iod scenario, for any given w
2 there is a threshold for w1 above/be-
low which the buyer will choose to use/not to use the
2 v Þðhþw2 cw1 Þ
: KðtÞ ¼ F1ðw2c
supplementary order (in their words, the system runs in the PAB/
w2vÞ y > 0;
Push regime). A counterpart observation for our problem is that,
is the unique Nash equilibrium for the stationary y K inventory game.

for any given w1 there is a threshold for w2 above/below which
the supplier will choose to provide/not to provide the supplemen-
w and w2 6 G(w1), then
tary supply. So our result provides a different dividing line in the
wholesale price space {w
1, w2} between the regimes that the sup-
plementary supply–order will and will not be activated, for a moregeneralized multi-period setting. Moreover, Dong and Zhu (2007)
is the unique Nash equilibrium for the stationary y K inventory game.

doesn't offer any implicit or explicit solution of the threshold theyannounce for w1. In contrast we derive the analytic expression forthis important boundary line in {w1, w2} map. Finally, in Dong and
Zhu (2007), the supply chain operating regimes depend on the de-
mand distribution, while in our paper Region 2 and Region 3 are
w2vÞ. If w1 >
w and w2 P w1, we
divided by w2 = G(w1), which do not depend on the speciﬁc de-
2 P w1 > G(w1) thus F1
mand distribution at all.

is also obtained. In this case, it is easy to verify that {y*(t), K*(t)}
The differences of the results stem from the fundamental differ-
expressed by (32) is a Nash equilibrium according to Lemma 5.

ence of decision interaction: they assume the buyer is a Stackel-
The initial inventory condition x0(0) = x(0) = 0 ensures that
berg leader in the supply–order game, so when w1 is small
xðtÞ < F1 ðw2vÞðrw1Þðw2cÞðrw2Þ
and XðtÞ < Sþ ¼ F1 w2c
enough the buyer will always choose to take full inventory risk;
for all t = 0, 1, . . , T 1, thus the above stationary base-stock/
our paper is based on the equal status setting between the supplier
base-system-stock policy can be proceeded and remains optimal
and the buyer, so even if w1 is very small, the supplier has the op-
for the buyer/supplier as period goes by. According to Theorem 1
tion to provide the supplementary supply for a large w2. Moreover,
this is the unique Nash equilibrium. Part (2) can be proved in a
most existent results in this paper are independent of the demand
similar manner and thus omitted.

distribution, which is also the direct result of the action order and
Theorem 2 states that the Nash equilibrium is also stationary
information set of the two players in our model. From a game the-
for the multi-period y K inventory game with stationary param-
oretic perspective, our paper is also an illustration where a slightly
eters over periods. Moreover, it provides a speciﬁc condition of
different interaction setting would lead to some very different
{w1, w2} under which the supplementary supply–order mechanism
takes effect. There exists a threshold value
w for the wholesale
Finally, we compare two special cases with the supplementary
price of normal order w1. If w1 is higher than this threshold, then
supply–order system. One is the multi-period selling-to-the-news-
any wholesale price of the supplementary order w2, as long as it
vendor system where no supplementary order is allowed. In this
is higher than w1, will motivate the supplier to actively offer a sup-
case, the buyer bears all the inventory risk and orders only once
plementary supply. See Region 1 in Fig. 1 for this case. If w1 is lower
in each period with wholesale price w1 before the demand is real-
than this threshold, the supplier would be induced to provide the
ized. The supplier produces exactly the order quantity and holds no
supplementary supply only at the wholesale price higher than
inventory. Since the parameters are stationary over the periods, the
1) (now just w2 > w1 is no longer enough to stimulate the sup-
buyer's optimal policy is order up to level y ¼ F1
plier). Note that G(w1) is decreasing in w1, which indicates that the
other case is the centralized system, i.e., the entire supply chain
lower w1 is, the higher w2 should be set to stimulate the supplier.

is controlled by a single manager and the total proﬁt is maximized.

See Region 2 in Fig. 1 for this case. In contrast, if both w1 and w2 are
In this case, w1 and w2 become trivial and the optimal policy for the
small, the supplier is not motivated to provide any supplementary
system is order up to level y ¼ F1 rc .

**Author's personal copy**

X. Li et al. / European Journal of Operational Research 209 (2011) 273–284
c 6 w1 6 w2 under a single-period setting. Theorem 3 extends this
inventory position
result into the multi-period situation with different decision inter-
action between the supply chain members. 3
the system optimum
4. Numerical studies
Service improvement
In this section we present several groups of numerical examples
to obtain more insights into the supplementary supply–order sys-
Inventory benefit
tem. To investigate the beneﬁt of the supplementary supply–order
*v*)(*r w *) (*w*
*c*)(*r w *)
opportunity, we mainly focus on the multi-period problem with
*v*)(*h w*
stationary parameters. We suppose that the demands are normallydistributed over the periods and the supplementary wholesaleprice w2 is large enough to stimulate the supplier (thus case 1 inTheorem 2 occurs).

For the supplementary
in the supplementary
in the selling to
supply-order system
centralized system
supply-order system newsvendor system
c = 0.95, and assume i.i.d variables D(t) are generated sto-chastically by a Normal distribution N(100, 30).

Fig. 2. Comparison on the base-stock inventory position among three systems.

The above two cases serve as benchmarks of the supplementary
Fig. 3 shows the relationship between the demands and order
supply–order system. We denote the supply chain proﬁt of the
quantities of the buyer with and without the supplementary order
multi-period selling-to-the-newsvendor system, supplementary
option. It indicates that the normal order quantity in the supple-
supply–order system, and centralized system as P , P, andP,
mentary supply–order system is smaller than that without the
respectively. Theorem 2 has shown that when w
supplementary order option, as the buyer has another opportunity
to replenish her inventory after the demand is realized.

2 6 G(w1), the supplementary supply–order system reduces to
the multi-period selling-to-the-newsvendor one. The opposite oc-
Fig. 4 shows the buyer's inventories in each period with and
casion, in which the supplementary mechanism takes effect, is
without the supplementary order option. It indicates that the sup-
compared with two benchmark cases as follows.

plementary order option reduces the buyer's inventory.

Fig. 5 shows the supplier's inventories in each period with and
Theorem 3. When w
without the supplementary order option. In the ‘‘selling to the
w and w2 > w1, or w1 6
w and w2 > G(w1),
newsvendor" case, the supplier does not hold any inventory. How-ever, in the supplementary supply–order system the supplier
(1) y > yðtÞ þ KðtÞ > y > yðtÞ,
keeps inventory and thus bears part of the demand risk. This risk
(2) P > P > P .

sharing between the two members in the supply chain is the basis
of the supplementary supply–order mechanism.

Figs. 6 and 7 show the improvement in performance when the
(3) P* increases in w2. When w2 ? r, we have yðtÞ þ KðtÞ ! y,
supplementary supply–order mechanism is utilized. It is clear that
and P ! P.

both the buyer and the supplier are better off as proﬁt increases astime passes.

Proof. When w1 >
w and w2 > w1, or w1 6
w and w2 6 G(w1), we
(2) Then we show the inﬂuence of price and cost parameters on
> yðtÞ. Fur-
w2v ; which implies yðtÞ þ KðtÞ > yn
the supplementary supply–order system. Normally distrib-
thermore, y > yðtÞ due to w
2 < r. Result (2) immediately follows
uted demand is again adopted, and the planning horizon T
Result (1). Finally, yðtÞ þ KðtÞ ¼ F1 w2c
is extended to 300. Each time we change one particular
to the result that P* is also increasing in w2.

parameter and ﬁxed the others, which have the same values
Theorem 3 shows that when the supplementary supply–order
as in the ﬁrst experiments.

mechanism takes effect, the base-stock inventory position andthe performance of the supply chain system are between the ones
More speciﬁcally, we investigate the parameter's sensitivity to
in the multi-period selling-to-the-newsvendor case (the worst
the proﬁt increment ratio. The proﬁt increment ratio is deﬁned
case) and the centralized case (the best case). More speciﬁcally,
as the proﬁt increment after allowing the supplementary supply–
the buyer's base-stock inventory position is lower than that in
order opportunity divided by the original proﬁt without the oppor-
the multi-period selling-to-the-newsvendor case, indicating that
tunity, which indicates the effect of the supplementary supply–or-
the supplementary mechanism helps the buyer to mitigate the de-
der mechanism.

mand variance and lower his inventory cost. However, the system
Table 2 summarizes Figs. 8–14 on the relationship between the
base-stock inventory position (i.e., the sum of the buyer's and sup-
parameters and the proﬁt increment ratio of the buyer, supplier
plier's) in the supplementary supply–order case is higher than that
and supply chain. In the table, ‘+'/‘' means the ratio is increas-
in the multi-period selling-to-the-newsvendor case, indicating that
ing/decreasing in the parameter, while ‘+' means ﬁrst increasing
the supplementary mechanism raises the service level for the sup-
then decreasing. We ﬁnd that from supply chain perspective, the
ply chain system. See Fig. 2 for a detailed illustration.

increase of cost c, price r and the supplier's holding cost h
On the other hand, both the system base-stock level and the
undermine the beneﬁt of supplementary supply–order mechanism
system proﬁt move closer to the centralized case as the supple-mentary wholesale price w2 increases towards its upper bound r,
the selling price. When w
Note that the single-period situation of our problem is a special case of our
2 arrives at r, the supply chain efﬁciency
stationary parameter situation, therefore all the results in Section 3.4, especially,
reaches 100%, i.e., the supplementary supply–order mechanism
Theorems 2 and 3, are still valid for the single-period supplementary supply–order
helps the supply chain achieve system optimum. Cachon (2004)
system. The only modiﬁcation we need is to change all cw1 h into v, which
states that the supply chain efﬁciency is 100% with w2 = r and
represents the salvage value of leftover product after that single period.

**Author's personal copy**

X. Li et al. / European Journal of Operational Research 209 (2011) 273–284
supplier's inventory with
supplementory supply
supplier's inventory without
supplementary supply
supplementory supply (always zero)
Fig. 3. Demands and buyer's orders.

Fig. 5. Supplier's inventory positions.

buyer's inventory with
supplementary order
buyer's inventory without
buyer's profit with
supplymentary order
supplementary order
buyer's profit withoutsupplementary order
buyer's profit improvement: 4.22%
Fig. 4. Buyer's inventory positions.

Fig. 6. Buyer's total proﬁt over multiple periods.

as the supplier's motivation of provide supplementary supply isweakened by the increase in these parameters. On the other hand,the increase of the buyer's holding cost h, wholesale prices w1, w2,
increase in the buyer's revenue but weakens the effect of the sup-
and the demand's uncertainty will enhance the effect of the sup-
plementary supply–order mechanism. The buyer's proﬁt incre-
plementary supply–order mechanism, since in these cases either
ment ratio is ﬁrst increasing and then decreasing in price r,
the supplier is more induced to provide a supplementary supply
which implies that the revenue increase compensates the weak-
(when w2 increases) or the buyer is more needful for the supple-
ness at ﬁrst, but ﬁnally the loss of effectiveness of the supplemen-
mentary order (when h, w1, r increases). In sum, given the same
tary supply–order mechanism becomes the dominant factor.

supply–order contract price, the supplementary supply–order sys-
Another interesting observation is that as w2 increases, the
tem is more beneﬁcial for a product with lower cost, lower value,
buyer's proﬁt increment ratio ﬁrst increases and then decreases.

lower inventory cost of supplier, higher inventory cost of buyer
This is because a higher w2 increases the buyer's cost since the sup-
and higher demand variation.

plementary order is more expensive than the normal order, mean-
We also notice that there are several cases in which the varia-
while it provides an incentive for the supplier to produce more
tion tendency of individual supply chain member is not consistent
units for the supplementary supply. When w2 is relatively small
with the system. For example, raising the selling price r incurs an
and increases, the beneﬁt of more supplementary supply over-
**Author's personal copy**

X. Li et al. / European Journal of Operational Research 209 (2011) 273–284
suppler's profit with
supplementory supply
suppler's profit without
supplementory supply
supplier's profit increment: 5.54%
profit increment ratio
Fig. 7. Supplier's total proﬁt over multiple periods.

Fig. 9. Inﬂuence of hs on the proﬁt increment ratio.

Table 2Relationship between parameters and proﬁt increment ratio.

Proﬁt increment ratio
profit increment ratio
profit increment ratio
Fig. 10. Inﬂuence of c on the proﬁt increment ratio.

supplier. From Fig. 14 we can see that the supplier's ratio ﬁrst in-creases but then zigzags when the variation in demand is larger.

It is interesting that in the above ﬁgures (Figs. 8–13), three trajec-tories always intersect at one point where the supply chain's curve
lies between the other two curves.

Fig. 8. Inﬂuence of h on the proﬁt increment ratio.

This paper presents a decentralized supply chain system with a
weights the side-effect of the increasing cost. However, when w2
supplementary supply–order opportunity under the multi-period
continues to increase and after a certain point, the increasing cost
situation. During each period the buyer places an order before
eventually becomes the dominant factor.

the demand is realized, and the supplier decides the production
Finally, the increase of w1 weakens the supplier's motivation to
quantity. The buyer might place a more costly supplementary or-
provide the supplementary supply thus reduces the beneﬁt for the
der after the uncertainty of the demand is resolved, and the order
**Author's personal copy**

X. Li et al. / European Journal of Operational Research 209 (2011) 273–284
profit increment ratio
profit increment ratio
Fig. 11. Inﬂuence of r on the proﬁt increment ratio.

Fig. 13. Inﬂuence of w2 on the proﬁt increment ratio.

the supplier chain
profit increment ratio 10%
profit increment ratio
standard deviation of demand: σ
Fig. 12. Inﬂuence of w1 on the proﬁt increment ratio.

Fig. 14. Inﬂuence of r on the proﬁt increment ratio.

can be satisﬁed by the supplier's supplementary supply. We adopt
loop Nash equilibrium is an important future research direction.

a game theoretic approach to analyze the model, propose the opti-
Moreover, in our study we assume that the selling price is exoge-
mal supply and order policies for the supplier and the buyer,
nous. Incorporating the pricing decisions into the model would
respectively, and show the existence and uniqueness of Nash equi-
make the problem more complicated but more interesting.

librium. Moreover, we obtain the analytical Nash equilibrium solu-tion when the parameters are stationary over the planning horizon.

It is shown that the mechanism, if properly designed, improves the
supply chain performance in the multi-period setting and beneﬁtsboth the supplier and the buyer.

The authors wish to express their sincerest thanks to the editors
Our paper is among the ﬁrst effort to study the multi-period
and anonymous referees for their constructive comments and sug-
supplementary supply–order game and the inventory decision
gestions on the earlier versions of the paper. We gratefully
interaction between the supply chain members. The equilibrium
acknowledge the support of (i) National Natural Science Founda-
concept is based on the information structure that each player con-
tion of China (NSFC), Research Fund Nos. 71002077 and
siders the strategy of his contender as static. Exploring the closed-
71002106, and 2009 Humanities and Social Science Youth Founda-
**Author's personal copy**

X. Li et al. / European Journal of Operational Research 209 (2011) 273–284
tion of Nankai University NKQ09027, for X. Li; (ii) NSFC Research
Choi, Tsan-Ming., 2007. Pre-season stocking and pricing decisions for fashion
Fund Nos. 70971069 and 71001106, and the Fok Ying-Tong Educa-
retailers with multiple information updating. International Journal ofProduction Economics 106 (1), 146–170.

tion Foundation of China No. 121078, for Y.J. Li; and (iii) Research
Deng, S., Yano, C., 2002a. On the Role of Second Purchase Opportunity in a Two-
Grants Council of Hong Kong, General Research Fund No. 410509,
Echelon Supply Chain. Working Paper, University of California, Berkeley.

and NSFC Key Program Grant No. 70932005, for X.Q. Cai.

Deng, S., Yano, C., 2002b. Combining Spot Purchases with Contracts in a Two-
Echelon Supply Chain. Working Paper, University of California, Berkeley.

Deshmukh, A., Lian, Z., 2009. Analysis of supply contracts with quantity ﬂexibility.

European Journal of Operational Research 196 (2), 526–533.

Appendix A. Supplementary data
Dong, L., Zhu, K., 2007. Two-wholesale-price contracts: Push, pull, and advance-
Management 9 (3), 291–311.

Supplementary data associated with this article can be found, in
Eppen, G.D., Iyer, A.V., 1997. Backup agreements in fashion buying – The value of
the online version, at doi:10.1016/j.ejor.2010.08.019.

upstream ﬂexibility. Management Science 43 (11), 1469–1484.

Fisher, M., Raman, M., 1996. Reducing the cost of demand uncertainty through
accurate response to early sales. Operations Research 44 (1), 87–99.

Fisher, M., Raman, M., 2001. Optimizing inventory replenishment of retail fashion
product. Manufacturing and Service Operations Management 3 (3), 230–241.

Ozer, O., Uncu, O., Wei, W., 2007. Selling to the newsvendor with a forecast update:
Anupindi, R., Bassok, Y., 1999. Supply contracts with quantity commitments and
Analysis of a dual purchase contract. European Journal of Operational Research
stochastic demand. In: Tayur, S., Magazine, M., Ganeshan, R. (Eds.), Quantitative
182 (3), 1150–1176.

Models for Supply Chain Management. Kluwer Academic Publishers., Norwell,
Puterman, M.L., 1994. Markov Decision Processes: Discrete Stochastic Dynamic
Programming. Wiley, New York.

Barnes-Schuster, Y., Bassok, D., Anupindi, R., 2002. Coordination and ﬂexibility in
Sethi, S.P., Yan, H., Zhang, H., 2001. Peeling layers of an onion: A periodic review
supply contracts with options. Manufacturing and Service Operations
inventory model with multiple delivery modes and forecast updates. Journal of
Management 4 (3), 171–207.

Optimization Theory and Applications 108, 253–281.

Bassok, Y., Anupindi, R., 1997. Analysis of supply contracts with total minimum
Sethi, S.P., Yan, H., Zhang, H., 2003. Inventory models with ﬁxed costs, forecast
commitment. IIE Transactions 29 (5), 373–381.

updates, and two delivery modes. Operations Research 51 (2), 321–328.

Brown, A.O., Lee, H.L., 2003. The impact of demand signal quality on optimal
Tsay, A.A., 1999. The quantity ﬂexibility contract and supplier–customer incentives.

decisions in supply contracts. In: Shanthikumar, G., Yao, D., Zijm, H. (Eds.),
Management Science 45, 1339–1358.

Stochastic Modeling and Optimization of Manufacturing Systems and Supply
Van Delft, C., Vial, J., 2001. Quantitative analysis of multi-periodic supply chain
Chains. Kluwer Academic Publishers., Boston, pp. 299–328.

contracts. working Paper, HEC, Paris. Available from <http://www.hec.fr/hec/fr/
Cachon, G.P., 2004. The allocation of inventory risk in a supply chain: push, pull, and
advance-purchase discount contracts. Management Science 50 (2), 222–238.

Wang, Q., Tang, O., Tsao, D., 2006. A ﬂexible contract strategy in a supply chain with
Chiang, C., 2003. Optimal replenishment for a periodic review inventory system
a ﬂexible production mode. International Journal of Operational Research 1 (3),
with two supply modes. European Journal of Operational Research 149 (1),
Yan, H., Liu, K., Hsu, A., 2003. Optimal ordering in a dual supplier system with
Chiang, C., 2010. An order expediting policy for continuous review systems with
demand forecast updates. Production and Operation Management 12 (1), 30–
manufacturing lead-time. European Journal of Operational Research 203 (2),
Zhao, Y., Wang, S., Cheng, T.C.E., Yang, X., Huang, Z., 2010. Coordination of supply
Chiang, C., Gutierrez, G.J., 1998. Optimal control policies for a periodic review
chains by option contracts: A cooperative game theory approach. European
inventory system with emergency orders. Naval Research Logistics 45, 187–
Journal of Operational Research 207, 668–675.

Zipkin, P.H., 2000. Foundations of Inventory Management. McGraw-Hill, New York.

Source: http://or.bsnku.net/papers/8%EF%BC%8Cejor-lilicai.pdf

PHARMA-BRIEF ESPECIAL À custa do pobre?Exame do comportamento empresarial da Boehringer Ingelheim, Bayer e Baxter no Brasil Membro da Health Action International I Brasil: país de contrastes . 1-13 O sistema de saúde brasileiro . 2-8 O mercado farmacêutico no Brasil . 8-10 As companhias examinadas . 11 II Métodos do estudo .14-19I I Resultados do estudo .20-45

GUIDELINES FOR THE ADMINISTRATION OF DRUGS THROUGH ENTERAL FEEDING TUBES 2nd Edition October 2003 NJWATT/PHARM/Review Date October 2004 Disclaimer Representations in this guide are believed to be true and accurate. The Primary Care Trusts, NHS Trusts, and their employees, or agents accept no liability for loss of any nature to persons, organisations, or institutions which may arise as a result of errors. Users of this guide are reminded that crushing medication or opening a capsule to aid administration results in unlicensed administration. It is recommended that unlicensed administration methods are authorised by the prescriber and documented in writing.