Int. J.

Production Economics 147 (2014) 108116

Contents lists available at ScienceDirect

Int. J. Production Economics

journal homepage: www.elsevier.com/locate/ijpe

Capacity planning and allocation with multi-channel distribution

Wenming Xie a,b, Zhibin Jiang a,b,n, Yingxue Zhao c, Junjie Hong c
a
Sino-US Global Logistics Institute, Shanghai Jiao Tong University, Shanghai, China
b
Department of Industrial Engineering & Management, School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China
c
School of International Trade and Economics, University of International Business and Economics, Beijing 100029, China

art ic l e i nf o a b s t r a c t

Article history: In this paper, we study a multi-channel distribution system in which a manufacturer sells its product via
Received 14 September 2012 an independent service provider and a direct selling market simultaneously. The manufacturer allocates
Accepted 7 June 2013 its production capacity to the service provider, and then the service provider supplements some value-
Available online 14 August 2013
added services to the product to satisfy the demand of the downstream customers. We assume that the
Keywords: manufacturer's production capacity planning and allocation decisions are challenged by information
Supply chain management asymmetry with the service provider and consider a two-stage model: at the rst stage, the manufacturer
Information asymmetry determines the optimal production capacity based on the service provider's order reservation and the
Multi-channel distribution demand forecast in the direct selling market. At the second stage, the manufacturer allocates the
Capacity planning and allocation
production capacity, taking into account updated demand from the service provider and the direct selling
market. In the paper, we propose several decision-making models and identify some structural properties
for them. In addition, policies for production capacity allocation between the service provider and the
direct selling market are developed with various different market scenarios. The numerical experiment
results show that introducing direct selling market can signicantly improve the system performance and
the decision-making mechanism developed in the paper is feasible and effective for the system.
& 2013 Elsevier B.V. All rights reserved.

1. Introduction production. However, since each single manufacturer cannot

Capacity planning is essentially important for effective strategic
decision-making in various industries. The examples can be found
in communication networks (Chang and Gavish, 1993; Ahlert et al.,
2009), electric utilities (Murphy et al., 1987), automobile industries
(Eppen et al., 1989), shipping industries (Wang and Yun, 2013;
Hung and Chen, 2013), and in electronic goods and semiconductor
industries (Chou et al., 2007; Chien et al., 2012). In all these
applications, the manufacturing capacity building requires a com-
mitment of substantial capital resource over a long period of time.
Furthermore, multi-channel distribution and high uncertainty
inherent in the long horizontal forecast for cost and demand make
the problems concerning capacity planning and allocation extre-
mely complex in supply chain management. One recent example
that can be shared is the following practical case from the fastener
manufacturing industry of China.
In the fastener manufacturing industry of China, there are more
than 10,000 types of product, and each type has several different
specications. In general, each downstream customer needs hun-
dreds types of fastener to meet the requirement of normal
n
Corresponding author at: Department of Industrial Engineering & Logistics
Management, School of Mechanical Engineering, Shanghai Jiao Tong University, 800
Dong Chuan, Shanghai 200240, China. Tel./fax: 86 21 34206065.
E-mail address: zbjiang@sjtu.edu.cn (Z. Jiang).
meet all demands of a single customer, the downstream customers
usually have to purchase the products from several manufacturers
at the same time, which brings about very high procurement cost
relative to the production cost. In order to change this operations
deciency, the manufacture sells the product to one distribution
center, and then the distribution center distributes all required
products to the customer with a customized manner, which helps
the customer reduce purchasing and inventory costs. In this new
operations pattern, the availability to all kinds of products and
services in an integrated way allows the manufacturer and the
distributor to form an alliance to create additional value sources
and thus gain competitiveness. However, there are also some
customers who do not need the distribution service from the
distribution center, and for these customers they can buy the
product directly from the manufacturer, which forms a direct
selling market. Actually, such business mode has become an very
popular practice in the manufacturing industries. Since the man-
ufacturers can sell their products via a common service provider
and a direct selling market simultaneously with such a business
mode, the optimal decisions on production capacity planning and
allocation needs a coordination between the demands from the
service provider and the direct selling market, which arises some
challenges for the manufacturer to well plan the production
capacity and its allocation prior to receiving orders from the
downstream partners.

0925-5273/$ - see front matter & 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.ijpe.2013.08.005
 W. Xie et al. / Int. J. Production Economics 147 (2014) 108116
With a purpose to resolve the issue mentioned above, this
paper deals with capacity planning and allocation in a dual-
channel supply chain, in which the manufacturer has production
capacity, and the downstream service provider has marketing
expertise and long-term relationships with the customers. The
service provider provides additional services that may attract
demands that are not directly accessible to manufacturers
(Gupta, 2008). Information asymmetry problem, existing in the
single channel supply chain, also exists in the multi-channel
supply chain. That is, the service provider has some private
information about the customers because it is closer to the
downstream market than the manufacturer. Besides the service
provider, the manufacturer can also sell the product via the direct
selling market. The demand from the direct selling market
uctuates considerably with volatility risk. Therefore, the manu-
facturer will make the production capacity decision incorporating
the demand information collected from both the service provider
and the direct selling market. In order to build a better manufac-
turing capacity, the manufacturer wishes to share the service
provider's private information as much as possible. However, the
service provider usually has an incentive to inform the manufac-
turer of an exaggerated demand, so that it can better avoid
capacity shortage in the expenditure of the manufacturer, which
leads to bullwhip effect in the supply chain. As a result, the
manufacturer often considers the demand information informed
by the service provider incredible. Therefore, it is necessary for the
manufacturer to develop an efcient decision-making mechanism
by which to encourage the service provider to share truthful
demand information with it, so that it can build an appropriate
production capacity. After the manufacturing capacity is built and
demand information is updated, the manufacturer will reconsider
the manufacturing capacity allocation proportion between the
service provider and the direct selling market. At this time point,
the manufacturer will prefer satisfying the demand from the direct
selling market to that from the service provider since the product
price is generally higher in the direct selling channel than in the
channel formed via the common distributor. Therefore, decision-
making mechanism needs to be designed and policy for produc-
tion capacity allocation between the direct selling market and the
service provider needs to be developed, taking into account
dynamic updation of the demand information, with the objective
to minimize the losses resulted from information asymmetry and
double marginalization effect in the decentralized supply chain.
In this paper, we propose a two-stage decision-making mechan-
ism to address the above incentive problem with multi-channel
distribution. The goal is to improve efciency in production capacity
planning and allocation, and at the same time achieve a coordinat-
ing assignment for the decision makers' incentives. In the paper, we
propose a principal-agent model for the capacity planning and
allocation problem. We develop an incentive scheme that can elicit
truthful private information of the service provider and ensure an
implementation of the optimal capacity planning that can max-
imize the manufacturer's expected prot. After demand information
updating, the manufacturer will reallocate the production capacity
between the service provider and the direct selling market. Since
market demand is usually uncertain, the updated demand from the
service provider is not necessarily equal to its reservation ordering
quantity, and hence the manufacturer needs to make a proper
decision on production capacity allocation between the service
provider and the direct selling market on an ad hoc basis. The rest
of the paper is organized as follows: Section 2 reviews the relevant
literature. Section 3 describes the problem under study and
formulates the model. Section 4 explores the optimal decisions for
manufacturing capacity planning and its reallocation. Section 5
presents numerical examples, and Section 6 concludes with sum-
mary of the paper.
109
2. Literature review
The problems concerning capacity planning and allocation in
the supply chain are challenging due to the long production lead-
time and high demand uncertainty. In order to highlight our
contribution, we rst review the literature that addresses different
aspects of the capacity planning and allocation. After that, we
review the literature that is particularly related to our work.
Capacity management in the supply chain has been a major
issue explored extensively in the literature (Van Mieghem, 2003).
Some useful approaches can be found in the literature that
addresses the participator's incentives in capacity management.
For example, Mallik and Harker (1998) design a rewarding func-
tion which can provide the incentive for marketing managers to
improve their forecast accuracy. Celikbas et al. (1999) develop a
penalty function to coordinate demand forecasting and production
planning. Porteus and Whang (1991) develop a transfer pricing
scheme to coordinate capacity planning and allocation for differ-
ent departments. Karabuk and Wu (2003) apply the approach
utilized in Makowski and Mezzetti (1994) to achieve the rst-best
prot in a semiconductor capacity allocation problem. Although
they also consider the issue of capacity contracting, they only
discuss this problem in supply chains with one manufacturer and
one supplier.
Another related research stream is the literature that addresses
issues concerning multi-channel distribution. Simchi-Levi et al.
(2004) provide a detailed review of this research stream. The
problem explored most for this research stream is the mutual
impacts of different channels. Chiang et al. (2003) study the issue
of multi-channel design. They show that the manufacturer can
gain a stable prot by introducing a direct selling channel. Yao and
Liu (2003) study the customer diffusion between a direct selling
channel and a traditional channel. They nd that demands in both
channels are stable under some conditions. Wallace et al. (2004)
indicate that multiple complementary channels can provide a
greater and deeper mix of customer service, and therefore enhan-
cing the seller's overall value proposition.
In summary, there is a substantial literature exploring the issue
of capacity planning and allocation in the supply chain. In most of
these studies, an inherent assumption is that all input parameters
in the decision models are available for decision maker. However,
this assumption fails when the decision maker has some private
information and are motivated to take advantage of it for local
gains. Although there is some literature exploring the capacity
planning and allocation problems with demand information
asymmetry, seldom of them considers this issue with multi-
channel distribution.
Hence, our paper contributes to the literature by the following
two aspects: rst, our paper develops a capacity reservation
contract for the supply chain, taking into account multi-channel
distribution and information asymmetry. The contract mechanism
developed in our paper can minimize the losses resulted from
information asymmetry and double marginalization effect. Sec-
ond, our paper explores the capacity reallocation problem caused
by the deviation between the forecasted demand and the realized
demand for four different market scenarios. We also explore how
to adjust the model parameters so as to reduce the deviation. The
numerical experiments presented in the paper demonstrate that
the results developed by us are effective in managing supply
chains with multi-channel distribution.
3. Problem description
The manufacturer needs to make two decisions before demand is
realized: capacity building and capacity allocation. The manufacturer
110 W. Xie et al. / Int. J. Production Economics 147 (2014) 108116

Table 1
The parameters (notations) used are as follows.

Variables

k The total capacity

The demand from the service provider's market. Is a continuous random variable with the mean value being zero. The corresponding
cumulative distribution function (c.d.f.) is G(.) and the probability density function (p.d.f.) is g(.)
The service provider's private information. It is not a common knowledge and the manufacturer has only a prior belief on , and the
corresponding cumulative distribution function (c.d.f.) is F(.) and the probability density function (p.d.f.) is f(.)
The demand in the direct selling market. Is a continuous random variable, and the corresponding cumulative distribution function
(c.d.f.) is Q(.) and the probability density function (p.d.f.) is q(.)
DS The forecasted demand in the direct selling market. Its updation is denoted as dS
DF The forecasted demands from the service provider, DF . Its updation is denoted as dF
The capacity proportion allocated to the service provider
h The capacity quantity allocated to the service provider, h k
l The payment transferred from the service provider to the manufacturer for capacity reservation
c Unit production cost
ck Unit capacity setup cost
w The price per unit of the product purchased by the service provider from the manufacturer
r The price that the service provider sells one unit product to its downstream customer
z The price that the manufacturer sells one unit product to the direct selling market

confronts the following difculties when making the decisions: the demand in the direct selling market is not correlated to the
(1) the demand information provided by the service provider may demand from the service provider.
be exaggerated, therefore it is difcult for the manufacturer to plan
The total demand comes from two sources: the demand DS
the total capacity; (2) the demand in the direct selling market is
from the direct selling market and the demand DF from the service
uncertain, it is therefore difcult for the manufacturer to allocate the
provider. In order to ensure protability for the service provider
capacity appropriately between the service provider and the direct
and the manufacturer, it should be required that w A c ck ; r
selling market. As the downstream partner, the service provider can
(Table 1). In addition, it is obvious that the direct selling price
decide the extent of sharing its private information with the
should be higher than the cost, i.e., z 4 c ck . Since the service
manufacturer with the purpose to optimize its prot. Therefore, a
provider provides additional services to customers based on the
capacity reservation contract needs to be developed to deal with this
manufacturer's product, the service provider's selling price should
problem. The proposed solution includes two stages: the capacity
be higher than the direct selling price, i.e., r 4 z. Given the capacity
planning stage (stage 1) and the capacity reallocation stage (stage 2).
k, when the central planner makes all the decisions, the total prot
At stage 1, the manufacturer provides a contract menu which consists
of the supply chain is
of the capacity allocation proportion and the payment for capacity
reservation of the service provider. After receiving the contract menu, T r  cE min ; k z  cE min; k1    ck k; 1
the service provider chooses an optimal contract based on its private
where the expectation is with respect to or . In the centralized
information including the forecast for customers' demand and its
supply chain, the service provider's private information is common
promotion effort. The manufacturer shares the private information of
knowledge, so is deterministically known to the manufacturer.
the service provider and builds an optimal production capacity. At
By setting the derivative of Eq. (1) to be zero, we have
stage 2, after new demand information is updated, the manufacturer
adjusts the parameters (the allocation proportion and the payment r  c c G c k  z  c1  c Q k1  c
for capacity reservation of the service provider) in the capacity r c c z  c1  c  ck
reservation contract, and the manufacturer may either compensate
the service provider for shortage or charge for additional capacity. let T C k r  c c G c k  z c1  c Q 1  c k, the optimal
Thus, the capacity allocation proportion between the service provider capacity in the centralized case is given by
and the direct selling market is further optimized and the payment k T C  1 r c c z  c1  c  ck :
c
2
for the service provider's capacity reservation is nally determined.

3.2. The wholesale price contract

3.1. Assumptions and notation
As described above, the manufacturer rst plans capacity and
then allocates the capacity to the service provider and the direct
selling market. Without loss of generality, the following assump-
tions are made in the paper:
Assumption 1. The manufacturer's capacity planning and alloca-
tion problem is studied in one period. The manufacturer can build
any size of capacity. The service provider cannot purchase product
from any other manufacturer.
Assumption 2. There are three players in the supply chain: the
manufacturer, the service provider, and the direct selling market.
All the prices are exogenous. The direct selling market and the
service provider's market do not belong to the same market. Thus,
Under wholesale price contract, the service provider pays a
wholesale price w to the manufacturer for each unit product.
Under the decentralized supply chain, the objectives of both the
manufacturer and the service provider are to maximize their own
prots. In this case, the manufacturer's and the service provider's
prots are as follow:
M w  cE min ; k z  cE min; k1   ck k; 3
F r  wE min ; k 4
with a decentralized supply chain, the manufacturer cannot share
the service provider's private information. To ensure that demand
is fully satised, the service provider may inform the manufacturer
of an exaggerated demand. The prot of the manufacturer can be
formulated as Eq. (3), and the service provider's demand DF is a
 W. Xie et al. / Int. J. Production Economics 147 (2014) 108116
random variable with DF . The manufacturer estimates the
cumulative distribution function and the probability density func-
tion of as F U and f U, respectively. By setting the derivative of
Eq. (3) to be zero, we have
w  c d F 3 G d k z  c1  d Q k1  d
w  c d z  c1  d  ck :
let T D k w  c d F 3 G d k z c1  d Q 1  d k. The optimal
capacity quantity can be calculated as
d
k TD 1 d d
w  c z  c1  ck ;
where F 3 G is the distribution function of . Comparing Eq. (2)
with the manufacturer's capacity decisions under symmetric and
asymmetric information in Eq. (5), we can see that capacity
d provider. The fourth term indicates the capacity setup cost. Eq. (7)
decision under the asymmetric information k is independent of
. Without credible sharing on the demand forecast information,
the manufacturer cannot adjust the capacity to response to the
service provider's demand forecast. As a result, both the manu-
facturer and the service provider will suffer a loss in prot. When
the service provider's private demand forecast is relatively high,
both parties will lose sales, which results in a decrease of the
prot. When the service provider's private demand forecast is
relatively low, the manufacturer will suffer a loss resulted from
excessive capacity. An effective way to deal with this inefciency is
to design contract mechanism by which to induce credible
information sharing in the supply chain.
4. Decision-making mechanism based on capacity planning
and allocation
Taylor and Xiao (2008) and Aviv (2007) study the problem of
how to assure credible information sharing in a suppliermanu-
facturer supply chain. They design some methods to make sure
that the manufacturer can benet from the information sharing.
However, those references only consider the information sharing
problem based on some particular cases like newsvendor or
scorecard model. In addition, they do not consider the adjustment
of the contract parameters. In our paper, we extend the informa-
tion sharing problem to the case with three partners and multi-
channel distribution.
The manufacturer's problem is to build production capacity and
allocate the production capacity. We consider a two-stage model
as follows: the rst is the capacity planning stage. At this stage, the
manufacturer designs a contract set fk; ; lg, which includes
the total capacity k, the capacity proportion allocated to the
service provider, and the payment l transferred from the service
provider to the manufacturer for capacity reservation. After
receiving the contract set fh; lg(which does not include the
values of satisfying h k), the service provider chooses the
~ l
optimal contract h; ~ based on its private information. After
that, the manufacturer observes the service provider's private
information  according to the selected contract h  ; l  ,
and builds the optimal capacity k  . Finally, the manufacturer
determines a payment l0 l  for the capacity reservation and
an initial capacity allocation h0 k  . The second is the
capacity reallocation stage. At this stage, demands from the service
provider and the direct selling market are updated. Based on the
total capacity k, the updated demand dF from the service provider,
and the updated demand dS in the direct selling market, the
manufacturer optimizes the contracting terms again. Finally, the
n
optimal n and l are determined by the manufacturer.
111
4.1. Stage 1: manufacturing capacity planning
The manufacturer designs a solution set denoted by
~ lg
fk; ; lg. Given a contract fh; ~ chosen, the manufac-
turer's and the service provider's prots can be expressed as
~ ; l;
M k; ~ w  cE min ; k
~
~ ~  ck k;
z  cE min; k1   l ~ 6


~ ; l;
F k; ~ r  wE min ; k
~ ~
 l; 7
where the expectation is with respect to or . Eq. (6) represents
5 the manufacturer's expected prot with the contract. The rst
term of it indicates the expected prot from the distribution
channel via the service provider. The second term indicates the
expected prot from the direct selling channel. The third term
indicates the revenue from the capacity reservation of the service
represents the service provider's expected prot with the contract.
The rst term of it represents the expected prot from selling the
product to the customers. The second term represents the pay-
ment to the manufacturer for the capacity reservation. From the
perspective of the manufacturer, it designs a solution set given by
fk; ; lg, which are transformed into a contract set fh; lg
from which the service provider chooses an optimal contract.
Hence, the manufacturer's problem is to solve
k; ; l argmaxf M k; ; l; g; 8
k;;l
Eq. (8) indicates that the manufacturer will try to optimize the
total capacity k, the capacity allocation proportion , and the
payment l for capacity reservation, so that it can obtain the
truthful private information of the service provider. From the
perspective of the service provider, it needs to decide the extent of
sharing its private information with the manufacturer based on
the reserved capacity h and the capacity reservation payment l.
Hence, the service provider's problem can be formulated as
h; l arg max f F h; l; g; 9
Eq. (8) shows that the service provider will choose an optimal
~ lg
contract fh; ~ and announce its private information to be ,
~
which may differ from the truthful private information .
4.1.1. Optimization of the contract parameters
From the manufacturer's perspective, it not only needs to
ensure a credible information sharing with the service provider,
but also provides an incentive for the participation of the service
provider. The problem can be formulated as follows:
maxE M k; ; l; 10
k;;l
s.t. IC: F k; ; l; Z F k; ~ ; l;
~ for all ~ a , PC:
F k; ; l; Z Fmin for all A ; .
The rst constraint set is to ensure incentive compatibility (IC),
the second constraint set is to ensure participation of the service
provider, which is called participation constraint (PC). The IC
encourages the service provider to reveal the truthful private
information and thereby ensures a credible information sharing
with the manufacturer. The PC means that the prot received by
the service provider with the participation is higher than that
without the participation and thereby ensures an incentive for the
service provider's participation. In addition, and denote the
upper and lower limits of , respectively. According to (10), we can
express the manufacturer's and service provider's prots, respec-
tively, as follow:
M M k; ; l; ; 11
112 W. Xie et al. / Int. J. Production Economics 147 (2014) 108116

F F k; ; l; : 12 It is obvious that the
objective function (10) can be rewritten as

R R
E M k; ; l; T k; ; l;  r  w Gkx  xdx
f d  min .
F

Lemma 1. Fmin F k; ; l; : By using integration by parts, the manufacturer's problem can

be formulated as
Z  
1  F
Proof. The service provider's optimal prot is F max~ F max T k; ;  r  wGk  f d  Fmin ;
k U ; f
~ ; l;
k; ~ . The envelope theorem implies that
15
d F ~ ; l;
d F k; ~
j~ r  wGk  13 let Hk; ; T k; ;  1  F=f r  wGk  .
d d
The rst term in Hk; ; is the total prot in the supply chain
since r  w 40 and Gk  4 0, it follows that F k; ; l; and the second term is the cost caused by information asymmetry.
is increasing in for all A ; . Hence, Fmin F k; ; l; . Hence, the manufacturer faces a tradeoff between maximizing the
total supply chain's prot and minimizing the service provider's
information rent. In a balance, the manufacturer will choose
Theorem 1. The contract set fk; ; lg is feasible if and only if the appropriate capacity and payment transfer to motivate the service
following conditions are satised: provider to reveal its truthful private information and at the same
Z time maximize the total surplus. Theorem 1 presents the feasible
1  1 F Fmin r  wGkx  xdx: 14 contract menu for the manufacturer. It means that when the
contract set fk; ; lg satises those two conditions given in
(1-2)k is increasing. Theorem 1, with the contract the manufacturer can ensure the
participation of the service provider and the service provider will
receive a maximum prot by revealing its private information
~ ; l;
~ = r  wGk ~  Z 0, if truthfully.
Proof. Since F k;
1 o 2 , F k~ 1 ; ; l~ 1 ; 1 r F k~ 1 ; ; l~ 1 ; 2 . Since the service Theorem 2. If F U has an increasing hazard rate andg U is increas-
provider with 2 maximizes the prot by revealing ~ 2 , we have r r
ing, then fk ; l g is the optimal solution of (10).
F k~ 1 ; ; l~ 1 ; 2 r F k~ 2 ; ; l~ 2 ; 2 ;
IC indicates that the service provider will achieve a maximum Proof. We rst nd the maximizer of Hk; ; for each The
prot when ~ i i . Thus, F k1 ; ; l1 ; 1 r F k2 ; ; l2 ; 2 . r
optimal capacity k can be solved by the rst order optimality
Therefore, the PC will be satised if F k; ; l; Z Fmin . condition, which is given as follows:
Furthermore, it is obvious that the manufacturer will increase l
r r
until the service provider's prot is just Fmin . Hence, the PC can be r  c1  Gk  1  z  c1 Q 1 k
replaced by PC : F k; ; l; Fmin . By integrating Eq. (13), 1  F r
 ck  r  wgk  0; 16
we can obtain the service provider's prot as F Fmin f
R
r  wGkx  xdx. It is easy to see the following results that: from Theorem 1, the necessary condition for the optimal solution
r
2 F k;;l;
r  wgk  4 0; is that k is increasing. By some calculations we obtain that
k
2 Hk;; r r
2 F k;;l;
 2 r  wgk  o 0; 2  2 r  cgk   1  2 z  cq1  k
k
2 k
1 F r
given 1 4 2 , if k1 o k2 , the following inequities will hold:  r  wgk ;
f
 
F k; ; l; 1 F k; ; l; 1 F k; ; l; 2 2 Hk; ; r d 1 F r
jk k1 4 jk k2 4 jk k2 ; r  cgk   r wgk 
k k k k d f
however, it is easy to see by the IC that F k;;l;1
jk k1 1  F r
k r  wgk  ;
F k;;l;2
jk k2 0. f
k
r 2
if gk  Z 0, then 2 Hk; ; =k r 0. Hence, Hk; ; is
concave in k. If F has an increasing hazard rate, then
Thus, a contradiction occurs, which implies that k1 Z k2 for r
d=d1  F=f r 0, 2 Hk; ; =k Z 0. Therefore, k
all 1 42 .
is increasing in .
In the following, we will prove those two conditions given in
Theorem 1 imply IC and PC. If , then PCis satised immedi-
ately. The service provider's prot can be expressed as r
Given the optimal solution k , we can obtain the service
Z ~ ; l;
~ xF provider's prot by (15) and derive the optimal payment by (8)
~ ; l;
~ k;
F k; dx to obtain that
x
Z r
l r  wE min ; k F : 17
~ ; l;
F k; ~ ~ ~  xdx
r  wGk
~
Z
F k; ; l; ~  x  Gkx  xdx
r wGk
c
Theorem 3. c k is the optimal reservation capacity for the
_ r c
centralized supply chain, r k o c k for all A ; .
~ since G U and k U are both increasing in , the integration
if 4 ,
in the above formulation is non-positive, and hence F k; ~
~
; l; r k; ; l; . In a similar way, we can show that this
F c
Proof. Since c and k are the optimal solutions for the
inequality holds for o . ~ Therefore, those two conditions given in r
centralized supply chain and r and k are the optimal solutions
Theorem 1 imply IC. for the decentralized supply chain with information asymmetry,

c r
we have T k ; c ; Z T k ; r ; . Moreover, r and k are
r c
the maximizer of Hk; ; , hence, we have Hk ; r ; Z Hk ; c ; .
Adding the two inequalities, we have
1  F
r c
r  wGk r  Gk c   Z 0:
f
r c
since G is increasing, it follows that r k r c k .
Theorem 3 shows that the service provider's reservation
capacity is affected mainly by its private information and the
optimal reservation capacity is less than that of the centralized
supply chain. Since the service provider's information rent F is
increasing in , the manufacturer will build a less capacity to
reduce the service provider's information rent. That is, there is a
downward distortion resulted from the information asymmetry.
4.2. Stage 2: manufacturing capacity reallocation
Although a substantial research has been conducted for the
issues concerning production capacity planning and allocation,
majority of these research is based on static decision rather than
dynamic decision. It is easy to see that, however, when the
demand forecast has an error as compared with the realized
demand, the effectiveness for contracts that are based on the
static decision will decrease. Gupta (2008) considers that capacity
contract is an essential component in the manufacturerservice
provider interactions. They nd that the exibility to adjust
contract parameter is relatively signicant. Actually, demand in
the supply chain is usually uncertain, and with new demand
information collected, it is necessary to explore how to reallocate
the capacity so as to further optimize the capacity allocation. In the
following, we develop a capacity reallocation model, taking into
account demand information updating.
n
With the total capacity k realized and new demand informa-
tion collected, the manufacturer will reconsider the allocation
proportion 0 at stage 2. If the capacity allocated to the service
provider at stage 2 is larger thanh, the manufacturer will charge an
additional fee from the service provider for the additional capacity
n n
l Z l0 , and vice versa l r l0 . Hence, the manufacturer's problem
can be formulated as
max M ; l
;l
s:t: F ; l Z F 0 ; l0 for all A 0; 1
the problem can be equivalently transformed as follows:
n
max M ; l maxw  cmindF ; k
;l ;l
n n
z cmindS ; k 1  l  ck k 
n n
s:t: r  wmindF ; k  l Z r  wmindF ; k 0  l0 for all A 0; 1;
by rearrangement, the above constraint can be rewritten as
n n
s.t. l r l0 r  w mindF ; k  mindF ; k 0  for all A 0; 1,it
is easy to see that the above constraint should be tight, thus the
manufacturer's problem becomes the following:
( n n
M
w  cmindF ; k z  cmindS ; k 1  l0
max max n n
r  w mindF ; k  mindF ; k 0   ck k
to summarize, the manufacturer's problem can be formulated as
follows:
n n
max M max fr  cmindF ; k  r  wmindF ; k 0
n n
z cmindS ; k 1  l0  ck k g:
W. Xie et al. / Int. J. Production Economics 147 (2014) 108116 113
r
Eq. (19) shows that the manufacturer will reconsider the
n
capacity allocation proportion based on the total capacity k , the
updated demand dF from the service provider, and the updated
demand dS in the direct selling market. In the following, we
categorize the market scenarios into four cases by the service
provider's demand forecast error and the manufacturer's capacity
size. Case 1.1: the service provider's demand forecast is lower than
the realized demand and the manufacturer has excessive capacity.
Case 1.2: the service provider's demand forecast is lower than the
realized demand and the manufacturer's capacity is insufcient.
Case 2.1: the service provider's demand forecast is higher than the
realized demand and the manufacturer has excessive capacity.
Case 2.2: the service provider's demand forecast is higher than the
realized demand and the manufacturer's capacity is insufcient.
When the capacity is insufcient at stage 2, the service
provider will suffer a loss resulted from shortage, which can be
measured by the compensation cost for backorder and order
cancellation. The compensation fee for backorder and order
cancellation is assumed to be exogenous and closely related to
the order expiration time and the bargaining power of the service
provider. Since the manufacturer is the predominant decision
maker in the supply chain, there is no compensation fee for
capacity shortage usually. It is the party who results in shortage
is responsible for compensating the customers. As to the manu-
facturer, the inventory cost resulted from overcapacity will lead to
a decrease in its prot. In the following analysis, we will omit this
term since it does not affect choice of the contract. We rst discuss
case 1
Case 1. the service provider's demand forecast is lower than the
n
realized demand, that is, dF Z k 0 . With this case, it is obvious that
Eq. (19) can be rewritten as
n n
max M maxr  cmindF ; k  r  wk 0
n n
z  cmindS ; k 1  l0  ck k ;
n n
to achieve a deal, the conditions k ZdF and k 1  Z dS should
hold. We dene these conditions as the optimal conditions of case
1.
Case 1.1. The manufacturer has excessive capacity, that is,
n
k Z dF dS . In this case, the manufacturer can satisfy both the
service provider's and the direct selling market's demands, that is,
the optimal conditions of case 1 hold. Due to demand forecast
error, the updated demand from the service provider is higher
than the capacity reserved at stage 1. Hence, the manufacturer will
increase the sales to the service provider and charge a fee for the
n
additional capacity. The manufacturer will set n dF =k to satisfy
all demand from the service provider, and its prot can be
expressed as
n n
M w  cdF z  cdS l  ck k ;
for simplicity, we set the ratio of the additional payment and the
n n
18 capacity reservation payment to satisfy l=l0 dF  k 0 =k 0 ,
n n
that is l dF  k 0 l0 =k 0 . Therefore, the total payment of the
n n n
service provider is l l0 dF  k 0 l0 =k 0 . In case 1.1, the man-
ufacturer's and the service provider's prots are given, respec-
tively, as follow:
) n n
M w  cdF z  cdS l0 dF  k 0 l0 =k 0 ck k ;
n
20
n :
n
F r  wdF  l0 dF  k 0 l0 =k 0 :
n
21
n
Case 1.2. The manufacturer's capacity is insufcient as k rdF dS .
In this case, the manufacturer cannot satisfy the total demands
from the service provider and the direct selling market, and hence
the optimal conditions of case 1 do not hold. Since the price is
19 higher in the direct selling market than at the service provider, the
114 W. Xie et al. / Int. J. Production Economics 147 (2014) 108116
manufacturer will rst satisfy the demand in the direct selling
market and then satisfy the demand of the service provider by the
remainder capacity. That is, the manufacturer will set n
n n
k  dS =k to satisfy the total demand in the direct selling market
and then sell the remainder to the service provider. Thus, its prot
n n n
is as follows: M w  ck  dS z  cdS l ck k .
In addition, in this case, the manufacturer will transfer a part of
the reserved capacity to satisfy the demand from the direct selling
market, so that it can increase the prot. However, the service
provider's updated demand can not be fully satised, even the
capacity reserved by it at stage 1 may not be fully satised. The
outcome resulted is that the service provider has to compensate
the customers for shortage. Meanwhile, the manufacturer has to
compensate the service provider so as to ensure that the
service provider's prot is not less than that of the case where it
receives the total capacity reserved. Hence, l  r w
n n
k 0  k  dS   , where denotes a part of the compensation
fee to the customers. One part of the compensation fee can be paid
by the manufacturer due to the shortage for the reserved capacity,
and the other part of it can be paid by the service provider due to
its forecast error. Therefore, the total payment of the service
n n n satisfy the excessive demand in the direct selling market, which
provider is: l l0  r  w k 0  k  dS   (without loss of
generality, it is assumed that the additional revenue from the
direct selling market is higher than the compensation fee incurred
by the service provider). In case 1.2, the manufacturer's and the
service provider's prots are as follow:
n n n n
M w  ck dS z  cdS l0  r  wk 0  k  dS    ck k ;
n n n
F r  wk  dS  l0 r  wk 0  k  dS   ; 23
where denotes the total compensation fee to the customers,
n n n
Z, and = k 0  k  dS =dF  k dS . Hence, the ser-
vice provider's prot function can be rewritten as
n dF  k  dS   .
F r  wk 0  l0   ;
Eq. (24) shows that the service provider can earn a prot as high as
that from the reserved capacity, and the loss resulted from its
forecast error is  .
Case 2. the service provider's demand forecast is higher than the
n
realized demand, that is, dF r k 0 . Similarly, Eq. (19) can be
rewritten as
n
max M maxfr  cmindF ; k  r wdF
n n
z  cmindS ; k 1  l0  ck k g:
n n
To achieve a deal, the conditions k Z dF and k 1  Z dS should
hold. We dene the two conditions as the optimal conditions
of case 2.
Case 2.1. The manufacturer has excessive capacity as k ZdF dS .
Using the arguments in a similar way to those in case 1.1, the
manufacturer can satisfy both the service provider's and the direct
selling market's demands. Therefore, the optimal conditions of
n n
case 2 hold. Hence, the manufacturer will set n A dF =k ; 1  dS =k 
in order to satisfy the demands from both the service provider and
the direct selling market. Thus, its prot can be expressed as
n n
M w  cdF z  cdS l  ck k ;
owing to the forecast error, the updated demand at the service
provider is lower than the capacity reserved. Hence, the service
provider will suffer a loss owing to over-reservation, since the
payment for the reserved capacity cannot reduce. In case 2.1,
the manufacturer's and the service provider's prots are given,
respectively, as follow:
n
M w cdF z  cdS l0  ck k ; 25
F r  wdF  l0 ; 26
n
Case 2.2. The manufacturer's capacity is insufcient as k r
dF dS . By a similar discussion to that in case 1.2, it is seen that
in this case the manufacturer cannot satisfy the total demands
from both the service provider and the direct selling market. Since
the optimal conditions of case 2 do not hold, the manufacturer
prefers to satisfy the demand from the direct selling market. That
n n
is, the manufacturer will set n k dS =k to satisfy the total
demand from the direct selling market, and then sell the remain-
der capacity to the service provider. Hence, its prot can be
formulated as
n n n
M w ck  dS z  cdS l  ck k :
Since the total capacity is insufcient and the demand in the
direct selling market is relatively large, the manufacturer prefers
to allocate some of the capacity reserved by the service provider to
will lead to that the demand from the service provider can not be
fully satised. Thus, there is a compensation fee to the customers.
Similarly, the compensation fee will be determined by bargaining
power of the service provider. Furthermore, the manufacturer will
make an optimal balance between the allocation proportion and
the compensation fee associated with shortage, with a purpose to
22 ensure that the prot received by the service provider is not
less than that when all demand is satised, that is, l  r 
n
wdF k  dS   , where denotes the compensation fee to the
customers. We assume the proportion between the compensation
fee in this case and that in case 1.2 is equal to the proportion
between the unsatised demand in this case and that in case 1.2.
n
Thus, the total payment of the service provider is: l l0  r  w
n
24
In case 2.2, the manufacturer's and the service provider's
prots are given, respectively, as follow:
 n
 n  n
M w  c k  dS z  cdS l0  r  w dF  k  dS   ck k ;
27
 n
 n 
F r  w k dS  l0 r  w dF  k  dS ; 28
where denotes the total compensation fee to the customers and
. The service provider's prot function can be rewritten as
F r  wdF  l0 ; 29
Eq. (29) shows that the service provider can obtain a total prot
resulted from the updated demand at stage 2, and the payment l0
for the reserved capacity keeps unchanged (note that due to the
error in demand forecasting, the capacity dF , allocated to the
n
service provider ultimately, is less than the reservation k 0 ).
n Since the manufacturer acts as the leader in the supply chain, it
can adjust the contract parameters so as to maximize the prot. By
employing the direct selling market, the manufacturer can
increase its market share and lower the risk of overcapacity. It is
relatively reasonable to assume that the supply chain member
should be responsible for the losses caused by itself. Hence, the
service provider will try to improve the demand forecast so as to
reduce its loss. In addition, the adjustment of contract parameters
should be able to ensure an efcient supply chain and a clear
assignment of the responsibilities and duties among the supply
chain members. The manufacturer has the responsibility to deter-
mine a reasonable capacity for the supply chain, and the service
provider is responsible for providing a timely and accurate report
about market demand forecasting.
 W. Xie et al. / Int. J. Production Economics 147 (2014) 108116 115

Table 2 Table 3
Comparison of the decisions and performances under different market conditions. Numerical analysis with four market scenarios.

Parameters Centralized supply Wholesale price Capacity reservation
chain contract contract
H 90 L 60 H 90 L 60 H 90 L 60
k 105.59 75.60 85.13 85.13 108.21 59.12
0.81 0.73 0.71 0.71 0.76 0.67
M 78.91 42.45 188.10 169.65
F 120.89 105.87 50.35 2.51
T 251.38 189.82 199.79 148.32 238.45 172.16
5. Example and managerial insight
5.1. Performances with wholesale price contract and capacity
reservation contract
In the example, we assume all cost and price parameters are
exogenous and w 8; r 10; z 9; c 2; ck 5;  U 50; 100,
 U  20; 20,  U 0; 60, and A 0; 1. The service provider
has a demand forecast which can not be observed by the
manufacturer, and the manufacturer can only estimate this piece
of information to be a random variable following a uniform
distribution over [50, 100]. We will conduct a comparison of the
capacity decision and prot under different operations patterns.
In Table 2, we conduct a comparison between the centralized
and the decentralized supply chains under two contracts with a
high demand forecast H 90 and a low demand forecast L 60,
respectively. With the comparison, we can see that: rst, the total
prot under the centralized system is higher than under the
wholesale price contract and under the capacity reservation
contract. Second, the manufacturer can always increase its own
prot as well as the system's prot by a capacity reservation
contract. This means that capacity reservation contract can provide
an improvement as compared with the wholesale price contract.
Third, the manufacturer's optimal capacity choice under the
wholesale price contract is constant and independent of the
service provider's private information. However, under the capa-
city reservation contract, the optimal capacity choice increases
from 59.12 to 108.21 when the demand forecast increases from 60
to 90. Fourth, the reservation capacity h k is closer to that of the
centralized case under a high demand forecast than under a
low demand forecast. The difference decreases from 15.58
(55.19  39.61) to 3.29(85.53  82.24) when demand forecast
increases from 60 to 90. This result is very consistent with
Theorem 3, which demonstrates that the manufacturer will build
a low capacity by which to reduce the service provider's
information rent.
5.2. Manufacturing capacity reallocation
When the demand forecast is 90, the manufacturer should
n Talents in UIBE. We would like to thank Na Li and Na Geng for their
build a capacity k 108:21 and allocate a proportion 0 0:76 to
n invaluable advice on the paper, and the editor and the anonymous
the service provider, that is, k 0 82:24 and the remainder
n referees for their constructive comments that have helped a lot in
allocated to the direct selling market is k 1  0 25:97 under
the capacity reservation contract. After that, with demand updat-
ing from the service provider and the direct selling market at stage
2, the manufacturer will reallocate the capacity to the service
provider and the direct selling market. In the example, by follow-
ing the cases discussed in Section 4.2, we demonstrate the
capacity allocation decision with four cases as follows: (1)
dF 90, dS 10, (2) dF 90, dS 30, (3) dF 80, dS 20, (4) dF 80,
dS 30. The corresponding optimal allocation proportion n , the
payment associated with the service provider's capacity reserva-
n
tion l , the compensation fee , the manufacturer's prot M , and
Cases Demand updating n l
n
M F
Case 1.1 dF 90, dS 10 0.83 0 120.34 189.47 0 59.66
Case 1.2 dF 90, dS 30 0.72 20.2 89.81 220.2 59.00 15.68
Case 2.1 dF 80, dS 20 [0.72, 0.82] 0 110.10 189.10 0 49.90
Case 2.2 dF 80, dS 30 0.72 9.00 97.50 235.70 9.00 49.90
the service provider's prot F are all shown in Table 3 (the unit
shortage reimbursement is assumed to be 5 in the example).
From Table 3, it is seen that when the updated demands are
dF 90 and dS 10, the optimal capacity allocation proportion is
n 0.83. When the updated demands are dF 90 and dS 30, the
optimal capacity allocation proportion is n 0.72, and the asso-
ciated compensation fee from the manufacturer to the service
provider is 20.2 and from the service provider to the customers is
59. When the updated demands are dF 80 and dS 20, the
optimal capacity allocation proportion n is in the range of from
0.72 to 0.82 and the associated compensation fee is zero. When
the updated demands are dF 80 and dS 30, there is a compensa-
tion fee from the manufacturer to the service provider for the
unsatised contract. The example in Table 3 demonstrates that the
results presented in Section 4.2 on contract parameter adjustment
are reasonable and effective.
6. Conclusion
In this paper, we explore the problem of capacity planning and
allocation in supply chains with information asymmetry and
multi-channel distribution. We also investigate the efciency of
the capacity reservation contract for the supply chain. In the
studies, a two-stage decision-making mechanism is developed to
improve the manufacturing capacity planning and allocation for
the supply chain. In addition, several practical models are pro-
posed and the corresponding properties are developed for them.
The numerical experiments show that the decision-making
mechanism developed by the paper is effective in solving the
manufacturing capacity planning and allocation problem.
Acknowledgments
This research was partially supported by the National Natural
Science Foundation of China (70932004, 71101028, and 71371052),
the Key Program of National Social Science Foundation of China
(13AZD006), the Specialized Research Fund for Doctoral Program
of Higher Education (20090073110035), the Program for New
Century Excellent Talents in University, the Research Fund for
Key Scientic and Innovation Project of Shanghai Municipal
Education Commission (09ZZ19), and the Program for Excellent
improving our paper.
References
Ahlert, K.H., Corsten, H., Gssinger, R., 2009. Capacity management in order-driven
production networksa exibility-oriented approach to determine the size of a
network capacity pool. Int. J. Prod. Econ. 118 (2), 430441.
Aviv, Y., 2007. On the benets of collaborative forecasting partnerships between
retailers and manufacturers. Manage. Sci. 53 (5), 777794.
Celikbas, M., Shanthikumar, J.G., Swaminathan, J.M., 1999. Coordinating production
quantities and demand forecasts through penalty schemes. IIE Trans. 31 (9),
851864.
116 W. Xie et al. / Int. J. Production Economics 147 (2014) 108116
Chang, S.G., Gavish, B., 1993. Telecommunications network topological design and
capacity expansion: formulations and algorithms. Telecommun. Syst. 1 (1),
99131.
Chiang, W.Y., Chhajed, D., Hess, J.D., 2003. Direct marketing, indirect prots: a
strategic analysis of dual-channel supply chain design. Manage. Sci. 49 (1),
120.
Chien, C.F., Wu, C.H., Chiang, Y.S., 2012. Coordinated capacity migration and
expansion planning for semiconductor manufacturing under demand uncer-
tainties. Int. J. Prod. Econ. 135 (2), 860869.
Chou, Y.C., Cheng, C.T., Yang, F.C., Liang, Y.Y., 2007. Evaluating alternative capacity
strategies in semiconductor manufacturing under uncertain demand and price
scenarios. Int. J. Prod. Econ. 105 (2), 591606.
Eppen, G.D., Martin, R.K., Schrage, L., 1989. A scenario approach to capacity
planning. Oper. Res. 37 (4), 517527.
Gupta, D., 2008. Flexible carrier-forwarder contracts for air cargo business. J.
Revenue Pricing Manage. 7 (4), 341356.
Hung, Y.F., Chen, C.H., 2013. An effective dynamic decision policy for the revenue
management of an airline ight. Int. J. Prod. Econ. 144 (2), 440450.
Karabuk, S., Wu, S.D., 2003. Incentive schemes for semiconductor capacity alloca-
tion: a game theoretic analysis. Prod. Operations Manage. 14 (2), 175188.
Makowski, L., Mezzetti, C., 1994. Bayesian and weakly robust rst best mechanisms:
characterizations. J. Econ. Theory. 64 (2), 500519.
Mallik, S., Harker, P.T., 1998. Coordinating Supply Chains with Competition:
Capacity-allocation in Semiconductor Manufacturing. Working Paper.
Fishman-Davidson Center, The Wharton School, University of Pennsylvania.
Murphy, F.H., Sen, S., Soyster, A.L., 1987. Electric utility expansion planning in the
presence of existing capacity: a no-differentiable, convex programming
approach. Comput. Operations Res. 14 (1), 1931.
Porteus, E.L., Whang, S., 1991. On manufacturing/marketing incentives. Manage. Sci.
37 (9), 11661181.
Simchi-Levi, D., Wu, D.S., Shen, Z.J., 2004. Handbook of Quantitative Supply Chain
Analysis: Modeling in the E-Business Era. Kluwer, Boston.
Taylor, T.A., Xiao, W., 2008. Does a manufacturer benet from selling to a better-
forecasting retailer? Manage. Sci. 56 (9), 15841598.
Van Mieghem, J.A., 2003. Capacity management, investment, and hedging: review
and recent developments. Manuf. Serv. Operations Manage. 5 (4), 269302.
Wallace, D.W., Giese, J.L., Johnson, J.L., 2004. Customer retailer loyalty in the context
of multiple channel strategies. J. Retailing 80 (4), 249263.
Wang, W.F., Yun, W.Y., 2013. Scheduling for inland container truck and train
transportation. Int. J. Prod. Econ. 143 (2), 349356.
Yao, D, Liu, J., 2003. Channel redistribution with direct selling. Eur. J. Oper. Res. 144
(3), 646658.