You are on page 1of 11

Int. J.

Production Economics xxx (xxxx) xxx

Contents lists available at ScienceDirect

International Journal of Production Economics


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

Efficient coordination of supply contracts with demand learning and two


production modes
Xin Li a , Zhaotong Lian b ,∗, Chuan Pang a
a
School of Business, Macau University of Science and Technology, Macau SAR, China
b
Faculty of Business and Administration, University of Macau, Macau SAR, China

ARTICLE INFO ABSTRACT

Keywords: We study a supply contract between a manufacturer and a retailer in a two-period newsvendor model with
Newsvendor demand learning. The retailer needs to determine the order quantity at the beginning, and can adjust the
Coordination amount at the end of the first period after receiving the updated demand information, as the demand forecast
Supply contract
for the second period (sales period) is dependent on the order quantity during the first period (reservation
Demand learning
period). Compared to the traditional buy-back contract, although the quantity adjustment (QA) can yield higher
Optimization
profits for both parties, the channel profit improvement by QA decreases with demand uncertainty.

1. Introduction likely to be outsourced to Asia (Ghemawat, 2003). Both basic items and
fashion items are provided in the sales period, but the fashion items
To reduce mismatch between supply and demand, many firms, such strategy offers flexibilities and allows firms to shift production quickly
as fashion companies, use two production modes: a slower planned when there is a supply disruption. Zara has become a classic case in
production mode in the first period, and a faster but typically more both the industrial and academic fields since 2000. Efficient sourcing
expensive production mode to produce a second run of products closer (e.g., sourcing from overseas) and responsive sourcing (e.g., sourcing
to the sales period. Under this two-mode strategy, retailers can take from a home country), studied by Wu and Zhang (2014), are similar
advantage of updated demand forecasts and adjust the order levels be- to the two items at Zara, as efficient sourcing usually provides a cost
fore the sales period. Donohue (2000) takes the publishing industry as advantage whereas responsive sourcing allows a firm to obtain more
an example that can benefit from the two-mode production. Books are accurate demand information for procurement decisions.
traditionally manufactured with long lead times because of large fixed As typical in seasonal fashion goods markets, although demand
costs in setting up printers and associated software. The demand for a uncertainty is high at the beginning of a season, there is a potential
particular book is often uncertain at the time of printing, especially if opportunity to learn about demand via early sales observations (Aviv
the book covers a trendy topic. Having a second production run is one
et al., 2019). Recently, a new marketing phenomenon has appeared in
way to eliminate inventory underage costs when the updated demand
the sale of electronic products, such as mobile phones. The life cycle
forecast exceeds expectation. From a channel perspective, this is a clear
of electronic products usually has two periods: the reservation period
win for the industry. More books are sold, more customers are satisfied,
and the sales period. Product retailers open reservations to customers
and potentially less inventory is wasted. Another similar case is Zara,
to test the demand before releasing new models to the market. During
in which about 40% of finished garments are manufactured internally,
the reservation period, the potential customers can receive production
and of the remainder, approximately two-thirds of the items are from
information online. For instance, The company Xiaomi announces a
Europe and North Africa, and one-third are from Asia. Zara classifies
the garments into two categories: basic items and fashion items. The new Xiaomi phone model. Xiaomi fans can reserve engineering/sample
fashion items with-a short lead time to produce and deliver tend to phones on-line and receive the phones after two weeks. The fans order
assume more business risks and therefore are either produced in small the new phones and not only act as consumers, but as opinion deliv-
lots internally or offered by suppliers according to the sales. For Zara, erers, idea creators, and sales representatives. Furthermore, the fans
the production price in Europe is typically 15%–20% more expensive write comments and disseminate product information on the internet.
than in Asia. Therefore, the production of basic items, that have a long A similar phenomenon is observed in many product and service online
lead time, and are more sensitive to price than time, is particularly sales. The merchants care about online comments and some merchants

∗ Corresponding author.
E-mail address: lianzt@um.edu.mo (Z. Lian).

https://doi.org/10.1016/j.ijpe.2019.107585
Received 3 March 2019; Received in revised form 5 November 2019; Accepted 11 December 2019
Available online 14 December 2019
0925-5273/© 2019 Elsevier B.V. All rights reserved.

Please cite this article as: Xin Li, Int. J. Production Economics, https://doi.org/10.1016/j.ijpe.2019.107585
X. Li et al. International Journal of Production Economics xxx (xxxx) xxx

even reward customers for positive comments. The fans who order dur- particular, we compare our results with those of the benchmarking
ing the reservation period have an influence on the buying behavior of model to show the impact of QA. We conclude the paper in Section 6.
later customers, and thus provide more accurate and updated demand All the proposition proofs can be found in Appendix A.
information to Xiaomi (Shih et al., 2014; Ma et al., 2015). We define the
above strategy as ‘‘reservation marketing’’. Based on the data analysis 2. Literature review
of the relationship between reservation quantity and demand in the
sales period, the retailer updates the demand forecast and modifies the There are two main research streams in the literature. One stream
order quantities at the beginning of the sales period. focuses on supply contracts with SCC, and the other focuses on demand
With the reservation marketing strategy, the parties in the supply learning in supply contracts.
chain are able to match supply with demand better by adjusting the
order quantity at the beginning of the sales period. Moreover, as this 2.1. Supply contracts with SCC
strategy consists of a reservation period and a sales period, the produc-
tion can be classified into planned production and adjusted production. We first review the literature on supply chain coordination. Spengler
Thus, the reservation marketing strategy can increase the supply chain’s (1950) mentions that independent, profit-maximizing firms extract less
overall profit. in a centralized supply chain managed by a single decision maker than
Motivated by the above innovation, we examine the impact of the in a decentralized supply chain, because of ‘‘double marginalization’’.
reservation quantity and the realization of the supply chain coordina- Hence, to realize effective management of the whole supply chain, the
tion. Although Cachon (2003) illustrates that many supply contracts supply chain coordination of activities is necessary among different
can realize the supply chain coordination (SCC), the contract that players in the channel through mechanisms such as return mechanisms
includes two wholesale prices and two buyback prices is more suitable (Pasternack, 2008), QF contracts (Tsay and Lovejoy, 1999), revenue
for a two-period newsvendor model. Under this type of contract, the sharing contracts (Cachon and Lariviere, 2005), or options (Barnes-
retailer can adjust (increase/decrease) the order quantity after the Schuster et al., 2002). A comprehensive review on SCC with contracts
reservation period. For example, Sun Microsystems, IBM, and Hewlett- can be found in Cachon (2003).
Packard have benefited from this type of contract with flexible order Supply chain coordination is an attractive topic that has been
quantity for components and assemblies since 1990 (see Tsay and studied by many scholars. Dockner and Fruchter (2014) introduce a
Lovejoy (1999)). dynamic transfer pricing scheme to coordinate operations and market-
The goal of this paper is to provide guidance for designing an ing departments. Dockner and Fruchter (2014) indicate that a dynamic
efficient pricing scheme that allows the channel to take full advantage transfer price can partially coordinates the two departments’ decentral-
of a second, faster production mode, and coordinating the channel with ized decisions without pre-commitment strategies, but can also fully
reservation-dependent demand. The demand distribution depends on coordinate decentralized decisions with pre-commitment. Through a
the reservation quantity in the reservation period. We first calculate perspective of conditional value at risk, Chen et al. (2014) study
the optimal quantity in the centralized system, in which the retailer and a decentralized supply chain including multiple risk-averse suppliers
the manufacturer can be considered as one party to maximize the total and a single risk-averse retailer. Because of the competition among
profit in the two-period newsvendor model. Using a supply contract suppliers, not all coordinating contracts are stable. Chen et al. (2014)
involving two wholesale prices and two buyback prices, we derive the introduce the notion of contract core and contract equilibrium to
profit functions of both the retailer and the manufacturer separately. In analyze the stability of coordinating contracts. Lei et al. (2015) present
this paper, we illustrate how to set an efficient price scheme between a newsvendor model to study the game process between a supplier that
these two parties to coordinate the supply chain. We also analyze the has private cost information and a retailer suffering from inventory
impact of demand uncertainty on the supply chain. The followings are inaccuracy problems. By setting the contract parameters appropriately,
the main contributions of this paper: both partners have the right incentive to maximize the total supply
chain profit: the supplier shares information actively by choosing the
1. Assuming that the demand distribution in the sales period can
contract designed for the marginal production cost, and the retailer’s
be updated with the conditional probability by using the or-
rational decision concurs with the overall optimal decision. Li et al.
der quantity of the reservation period, we derive the retailer’s
(2016) study a QF contract between a cosmetic manufacturer and
optimal order policy for the sales period, including the order
a retailer with SCC in two stages. Saha and Goyal (2015) discuss
quantity at the beginning of the reservation period, and the ad-
some SCC contracts in which the demand is dependent on inventory
justment quantity at the beginning of the sales period. Compared
level and retail price. Weraikat et al. (2016) study the coordination
to the traditional buy-back contract, in this contract that uses
in a two-echelon pharmaceutical reverse supply chain with customer
the reservation marketing strategy, the supply better matches
incentives. Other scholars consider three-echelon SCC with social re-
the demand and the supply chain is coordinated. As a result,
sponsibility (Panda et al., 2015), loss-averse retailers (Hu et al., 2016),
the parties in the supply chain benefit from both SCC and the
and duopolistic retailers with perfect quality products (Modak et al.,
quantity adjustment. Consequently, the parties yields higher
2015). However, in the above mentioned studies, demand learning is
profits and the Pareto improvement can be achieved.
not involved.
2. Henig et al. (1997)’s conclusion that the thresholds of the ad-
justment are constants, we find that the thresholds vary with
2.2. Demand learning in the supply contract
different reservation quantities. We also find that the quantity
adjustment can yield higher channel profit than that of the buy-
Most of the above mentioned studies assume that the demand
back contract. However, the increment of the channel profit
forecast is not updated at the beginning of the sales period. Below
strictly decreases with demand uncertainty. Using the quantity
we review the literature on SCC with demand information updating.
adjustment, the improvement in the supply chain performance
There are several approaches to modeling demand learning: Bayesian
increases and then decreases with the marginal profit.
models (Dvoretzky et al., 1952; Eppen and Iyer, 1997; Choi et al.,
The structure of this paper is as follows. We review the related 2003), time-series models (Johnson and Thompson, 1975; Lovejoy,
literature in Section 2. In Section 3, we describe the basic model and 1990), and forecast revisions (Hausmann, 1969; Sethi and Sorger, 1991;
introduce the demand distribution updating. We explain how to apply Heath and Jackson, 1994; Donohue, 2000; Yan et al., 2003; Gurnani
the contract to achieve SCC in Section 4. In Section 5, we provide and Tang, 1999; Gallego and Özer, 2001). A more detailed review on
the results of the sensitive analysis and offer managerial insights. In demand information updating can be found in Sethi et al. (2006). Sethi

2
X. Li et al. International Journal of Production Economics xxx (xxxx) xxx

et al. (2004) study single- and multi-period quantity flexibility contracts


that involve one demand forecast update in each period in a spot
market. Many scholars study the pricing and signaling strategy with
strategic customers in multi-period settings. Yu et al. (2014) consider
a seller who can sell her product over two periods. They characterize
the seller’s signaling strategy and find that the rationing of capacity
in the advance period is an effective tool to signal product quality.
They find that a high-quality seller can distinguish herself by allocating
less capacity than the low-quality seller in the advance period. Yu
et al. (2015a) also study the impact of consumer-generated quality
information on a firm’s dynamic pricing strategy in the presence of
strategic consumers. They also use many examples to show that the
informativeness of consumer-generated quality information depends on
the volume of consumers who share their opinions and the initial sales
volume. Aviv et al. (2019) study the potential benefits of responsive
pricing and demand learning to sellers of seasonal fashion goods. They
find that the benefits of responsive pricing, compared to those of a
fixed-price policy, depend sharply on the nature of the consumers’
behavior. Specifically, in contrast to markets of myopic consumers,
when the consumers are all strategic, the benefits of responsive pricing
tend to diminish when there is a higher potential for learning. Similar
studies can be found in Yu et al. (2015b) and Wei and Zhang (2018).
The above literature mainly focuses on the trading between the retailer
Fig. 1. Reservation-dependent Demand Model.
and the final consumers, and as such the retail price is an endogenous
variable. However, our work focuses on the ordering policy between
the manufacturer and the retailer, so the retail price is an exogenous
variable. 3. Model
Many scholars study ordering policy in multiple periods through
demand signals or demand learning. Wu and Zhang (2014) study We consider a supply contract with demand learning, in which a
a sourcing game in which competing firms can choose between effi- retailer and a manufacturer reach consensus in a two-period newsven-
cient sourcing (e.g., sourcing from overseas) and responsive sourcing dor model. Without loss of generality, we use the following example
(e.g., sourcing from a home country). Hu et al. (2015) formulate to illustrate this model. Consider that a new type of mobile phone is
and analyze a novel model of a firm’s dynamic inventory and mark- commercialized. Before the sales period, to test the market, the retailer
down decisions for perishable goods in two phases of every period: allows consumers to reserve phones on line. 𝑋 denotes the total amount
the clearance phase and the regular-sales phase. The firm needs to of the reservation. The cumulative distribution function (CDF) of 𝑋
make a trade-off between product spoilage and intertemporal demand is assumed to be 𝐹 (𝑋) and the probability density function (PDF) is
substitution. In their study, Hu et al. (2015) find that the firm should assumed to be 𝑓 (𝑋). 𝑌 denotes the total demand in the sales period.
either put all of the leftover inventory on discount or dispose of it, At the beginning of the sales period, the reservation quantity 𝑋 = 𝑥
with the choice depending on the amount of leftover inventory from is known, and the demand distribution in the sales period can be
the previous period. estimated. Then, the PDF of 𝑌 can be expressed as ℎ(𝑦|𝑥).
In our study, we assume that the reservation quantity also affects
The corresponding CDF is
the customer demand in the sales period and the retailer’s decision.
𝑦
Therefore, we use the conditional probability to update the demand 𝐻(𝑦|𝑥) = ℎ(𝑧|𝑥)𝑑𝑧, 𝑦 ≥ 0. (1)
information at the beginning of the sales period. We conduct a supply ∫0
contract to coordinate the channel profit in two periods and realize Based on the law of total probability, the CDF of 𝑌 at the beginning of
the Pareto improvement for the manufacturer and the retailer in the the reservation period (when 𝑋 is unknown) is
channel. +∞
Our work is closely related to that of Dong and Zhu (2007) in which 𝐻(𝑦) = 𝐻(𝑦|𝑥)𝑑𝐹 (𝑥), 𝑦 ≥ 0. (2)
a retailer is offered two ordering opportunities before the selling period- ∫0
i.e., an early order is allowed before the supplier’s production decision, Given the reservation quantity, the retailer can estimate the sales pe-
and a late order is allowed even after the completion of production. riod demand more accurately and adjust the order quantity. Below we
Dong and Zhu (2007) adopt a unified two-wholesale-price contract see how to generate a supply contract to coordinate the manufacturer
framework to study the Pareto improvement opportunities available to and the retailer.
systems using push, pull or advance purchase discount contracts. More By analyzing the effect of the reservation quantity on the demand
importantly, the authors identify the effect of the wholesale pricing in the sales period, we can better determine the reservation quantity,
on the inventory decision rights and ownership in the absence of the
which thus provides more accurate, updated demand information to
‘‘usual’’ reasons for advance-purchase discounts such as information
the retailers for the sales period. As shown in Fig. 1, we assume
asymmetry and promotion/information acquisition efforts. Our model
that customers who make reservations can create more demand. The
is distinct from Dong and Zhu (2007)’s model in several ways: (1)
distributions of 𝑋 and 𝑌 are both general.
the demand in the reservation period plays a role in updating the
demand information, and the reservation items are delivered in the
second period; (2) at the beginning of the second period, the retailer 3.1. Model description
can adjust the order quantity by either reducing the quantity (lose the
deposit) or ordering extra items with a higher price (see Li et al., 2016); The relationship between the manufacturer and the retailer is de-
(3) in the case that the demand in the reservation period plays a role in scribed in Fig. 2.
updating the demand information, the supply contract can arbitrarily The manufacturer and the retailer achieve a commitment in which
split the channel profit between the manufacturer and the retailer, and the retailer commits an order with a wholesale price 𝑤0 at the be-
coordinate the supply chain at the same time. ginning of the horizon. Assume that the order quantity is 𝑄. The

3
X. Li et al. International Journal of Production Economics xxx (xxxx) xxx

Fig. 2. Illustration of the model.

manufacturer’s production cost is 𝑚0 per unit. During the first period Recall that 𝑋 is the reservation quantity, and 𝑌 is the total demand
(the reservation period), the final customers can reserve the product in the sales period. Therefore, the total revenue consists of two terms:
with a price 𝑝1 and the retailer delivers the items to the customers 𝑝1 𝑋 + 𝑝2 min{𝑄̃ − 𝑋, 𝑌 }. The production cost also consists of two terms:
at the end of the first period. Observing the reservation quantity at 𝑚0 𝑄 + 𝑚1 (𝑄̃ − 𝑄)+ . The first term 𝑚0 𝑄 is the cost in the first period and
the end of the first period, the retailer has an updated estimate of the the second term 𝑚1 (𝑄̃ − 𝑄)+ is the cost in the second period. The items
demand for the second period. 𝑄̃ denotes the order quantity after the returned at the end of the first period save 𝑠1 (𝑄̃ − 𝑄)− , and the items
adjustment, where 𝑄̃ can be greater than or less than 𝑄. If 𝑄̃ ≥ 𝑄, the unsold at the end of the second period save 𝑠2 (𝑄̃ − 𝑋 − 𝑌 )+ .
manufacturer will charge a higher price 𝑤1 (𝑤1 ≥ 𝑤0 ) for each extra In the channel profit function, there are two decision variables: 𝑄,
amount 𝑄̃ − 𝑄. Producing the products also costs the manufacturer an the production quantity decided at the beginning of the first period
expedited unit cost 𝑚1 (𝑚1 ≥ 𝑚0 ) because of the shorter lead time. If and 𝑄,̃ the updated production quantity after the adjustment at the
𝑄̃ ≤ 𝑄, the manufacturer pays a buy-back price 𝑏1 to the retailer for beginning of the second period. We use dynamic programming to
the returned amount 𝑄 − 𝑄. ̃ derive the optimal solutions of 𝑄 and 𝑄. ̃
During the second period (the sales period), the retailer immediately Let 𝐸𝑋 and 𝐸𝑌 be the expectations of the items in braces, with
deliveries the items to the customers who order after the reservation random variables 𝑋 and 𝑌 respectively. The expected channel profit
period. The new customers pay price 𝑝2 . function is expressed as
At the end of the second period, the unsold items are bought back
by the manufacturer at buy-back price 𝑏2 . The manufacturer gets back ̃
𝜋𝐺 = 𝑝1 𝐸𝑋 {𝑋} − 𝑚0 𝑄 + 𝐸𝑋 {max 𝑈𝐺 (𝑄|𝑋 = 𝑥)}, (4)
𝑄̃
a salvage value 𝑠1 per item if it is returned at the end of the first
period, and 𝑠2 per item if it is returned at the end of the second period. where
For convenience, we name ‘‘quantity-adjustment contract with demand ̃
𝑈𝐺 (𝑄|𝑋 = 𝑥) = 𝑝2 𝐸𝑌 {min{𝑄̃ − 𝑥, 𝑌 |𝑥}} − 𝑚1 (𝑄̃ − 𝑄)+
learning’’ as the QACDL contract.
Thus, we have the following reasonable assumption: 𝑝2 ≥ 𝑤1 ≥ 𝑤0 ≥ + 𝑠1 (𝑄̃ − 𝑄)− + 𝑠2 𝐸𝑌 {(𝑄̃ − 𝑥 − 𝑌 )+ |𝑥}.
𝑏1 ≥ 𝑠1 . To motivate the retailer to modify the order quantity if needed, We maximize the expected supply chain profit by using a backward
we assume 𝑏1 > 𝑏2 , 𝑝2 > 𝑤1 , 𝑚0 > 𝑠1 and 𝑚1 > 𝑠2 . We also assume induction method. In the second period, given 𝑄 and 𝑋 = 𝑥, by
𝑤1 − 𝑏1 < 𝑚1 − 𝑠1 such that the retailer and the manufacturer share the ̃
maximizing 𝑈𝐺 (𝑄|𝑋), we first determine 𝑄̃ ∗𝐺 , the optimal adjusted
risk of the unsold products. ̃ Hence 𝑄̃ ∗ is a function of 𝑄 and 𝑥. In the first period, by
quantity of 𝑄. 𝐺
maximizing 𝜋𝐺 (𝑄), we determine 𝑄∗𝐺 , the optimal production quantity
4. The model with general demand distribution
of 𝑄.
We have the following Lemma.
We first consider the supply chain as a centralized system-i.e., we
assume that the manufacturer and the retailer are one party. Through
Lemma 1. Given 𝑄 and 𝑋 = 𝑥,
a backward induction method, we can derive the optimal production
quantity and the adjusted production quantity to maximize the ex- ⎧ 𝑆 𝐿 (𝑥) + 𝑥, if 𝑥 ≥ 𝑥𝑈 ,
pected channel profit. We illustrate below how to coordinate the supply ̃ ∗ ⎪ 𝐺 𝐺
𝑄𝐺 = ⎨ 𝑄, if 𝑥𝐿
𝐺
≤ 𝑥 < 𝑥𝑈
𝐺
, (5)
chain to obtain the Pareto improvement (a win–win outcome for the ⎪ 𝑆 𝑈 (𝑥) + 𝑥, if 𝑥 < 𝑥𝐿
two parties). ⎩ 𝐺 𝐺
( ) ( )
𝑝2 −𝑚1 𝑝2 −𝑠1
where 𝑆𝐺𝐿 (𝑥) = 𝐻 −1 𝑝2 −𝑠2
|𝑥 , and 𝑆𝐺𝑈 (𝑥) = 𝐻 −1 𝑝2 −𝑠2
|𝑥 . 𝑥𝑈
𝐺
solves
4.1. Centralized system
𝑄 = 𝑆𝐺𝐿 (𝑥) + 𝑥, (6)
In the centralized system, we derive the optimal production quantity
of 𝑄 at the beginning of the first period and the optimal adjusted and 𝑥𝐿
𝐺
solves
production quantity of 𝑄̃ at the beginning of the second period by
maximizing the expected channel profit. 𝑄 = 𝑆𝐺𝑈 (𝑥) + 𝑥. (7)
As 𝑤𝑖 and 𝑏𝑖 , 𝑖 = 1, 2 are the transfer prices-the price of the goods Consequently, 𝑆𝐺𝐿 ≤ 𝑆𝐺𝑈 and 𝑥𝐿 ≤ 𝑥𝑈 .
𝐺 𝐺
sold between the manufacturer and the retailer, they are not included
in the expected channel profit.
Remark 1. When the reservation quantity 𝑥 is less than the threshold
At the beginning of the second period, if 𝑄̃ ≥ 𝑄, the production cost
̃ 𝑥𝐿 , the adjusted quantity 𝑄̃ ∗𝐺 should be reduced to 𝑆𝐺𝑈 (𝑥) + 𝑥. When the
of the extra 𝑄−𝑄 are 𝑚1 per item. In contrast, if the 𝑄̃ ≤ 𝑄, the salvage 𝐺
reservation quantity 𝑥 is greater than the threshold 𝑥𝑈 , the adjusted
value 𝑠1 per item will be saved at the end of the first period. 𝐺
quantity 𝑄̃ ∗𝐺 should be increased to 𝑆𝐺𝐿 (𝑥) + 𝑥. When the reservation
In general, the channel profit is expressed as:
quantity 𝑥 is between 𝑥𝐿 𝐺
and 𝑥𝑈
𝐺
, the order quantity need not be
Channel profit = Total revenue - production cost + salvage value. (3) adjusted.

4
X. Li et al. International Journal of Production Economics xxx (xxxx) xxx

The thresholds 𝑆𝐺𝐿 (𝑥) and 𝑆𝐺𝑈 (𝑥) are not constants if the distribution where
of demand is dependent on the reservation quantity. In other words, ̃
𝑈𝑅 (𝑄|𝑋 = 𝑥) = 𝑝2 𝐸𝑌 {min{𝑄̃ − 𝑥, 𝑌 }|𝑥} − 𝑤1 (𝑄̃ − 𝑄)+
the thresholds of the adjusted production quantity (𝑆𝐺𝐿 and 𝑆𝐺𝑈 ) change
with 𝑥. This is very different from the result in Henig et al. (1997) in + 𝑏1 (𝑄̃ − 𝑄)− + 𝑏2 𝐸𝑌 {(𝑄̃ − 𝑥 − 𝑌 )+ |𝑥}.
which the thresholds are constant over the whole horizon. As shown ̃ we obtain the optimal value 𝑄̃ ∗ of 𝑄,
By maximizing 𝑈𝑅 (𝑄), ̃
𝑅
in (40), the demand increases in the reservation quantity (𝑥). Conse-
quently, both 𝑆𝐺𝑈 (𝑥) and 𝑆𝐺𝐿 (𝑥) increase with 𝑥. Coupled with (6) and ⎧ 𝑆 𝐿 (𝑥) + 𝑥, if 𝑥 ≥ 𝑥𝑈 ,
(7), both 𝑥𝑈 and 𝑥𝐿 are unique. ̃ ∗ ⎪ 𝑅 𝑅
𝐺 𝐺 𝑄𝑅 = ⎨ 𝑄, if 𝑥𝐿
𝑅
≤ 𝑥 < 𝑥𝑈
𝑅
, (11)
⎪ 𝑆 𝑈 (𝑥) + 𝑥, if 𝑥 < 𝑥𝐿 ,
Coupled with (5), the expected channel profit is ⎩ 𝑅 𝑅
∞ ( ) ( )
𝑝2 −𝑤1 𝑝2 −𝑏1
𝜋𝐺 = 𝑝1 𝑥𝑑𝐹 (𝑥) − 𝑚0 𝑄 where 𝑆𝑅𝐿 (𝑥) = 𝐻 −1 𝑝2 −𝑏2
|𝑥 , and 𝑆𝑅𝑈 (𝑥) = 𝐻 −1 𝑝2 −𝑏2
|𝑥 . 𝑥𝑈
𝑅
solves
∫0
{[ 𝐿 (𝑥)
]
∞ 𝑆𝐺 ∞
𝑄 = 𝑆𝑅𝐿 (𝑥) + 𝑥, (12)
+ 𝑦𝑑𝐻(𝑦|𝑥) + 𝑆𝐺𝐿 (𝑥)𝑑𝐻(𝑦|𝑥) 𝑝2
∫𝑥𝑈 ∫0 ∫𝑆 𝐿 (𝑥)
𝐺
}
𝐺 and 𝑥𝐿
𝑅
solves
𝐿 (𝑥)
𝑆𝐺 ( 𝐿 )
+ 𝑠2 𝑆𝐺 (𝑥) − 𝑦 𝑑𝐻(𝑦|𝑥) 𝑑𝐹 (𝑥) 𝑄 = 𝑆𝑅𝑈 (𝑥) + 𝑥. (13)
∫0
𝑥𝑈 {[ 𝑄−𝑥 ∞ ] Consequently, ≤ 𝑆𝑅𝐿and 𝑆𝑅𝑈 ≤ 𝑥𝐿
𝑅
𝑥𝑈
𝑅
.
𝐺
+ 𝑦𝑑𝐻(𝑦|𝑥) + (𝑄 − 𝑥)𝑑𝐻(𝑦|𝑥) 𝑝2 Given the prices, the retailer’s objective is to obtain the optimal
∫𝑥𝐿 ∫0 ∫𝑄−𝑥
𝐺 commitment quantity of 𝑄 by maximizing the retailer’s profit. We have
𝑄−𝑥 }
the following proposition.
+ 𝑠2 (𝑄 − 𝑥 − 𝑦) 𝑑𝐻(𝑦|𝑥) 𝑑𝐹 (𝑥)
∫0
𝐿
{[ 𝑈
] Proposition 4.2. The retailer’s optimal order quantity 𝑄∗𝑅 is the unique
𝑥𝐺 𝑆𝐺 (𝑥) ∞
+ 𝑦𝑑𝐻(𝑦|𝑥) + 𝑆𝐺𝑈 (𝑥)𝑑𝐻(𝑦|𝑥) 𝑝2 solution of the equation:
∫0 ∫0 ∫𝑆 𝑈 (𝑥)
}𝐺
[𝑝2 − 𝑏1 ]𝐹 (𝑥𝐿 𝑈
𝑈 (𝑥)
𝑆𝐺 ( 𝑈 ) 𝑅 (𝑄)) − [𝑝2 − 𝑤1 ]𝐹 (𝑥𝑅 (𝑄)),
+ 𝑠2 𝑆𝐺 (𝑥) − 𝑦 𝑑𝐻(𝑦|𝑥) 𝑑𝐹 (𝑥) 𝑥𝑈 (𝑄)
∫0 + (𝑝2 − 𝑏2 )
𝑅
𝐻(𝑄 − 𝑥|𝑥)𝑑𝐹 (𝑥) = 𝑤1 − 𝑤0 . (14)
∫𝑥𝐿 (𝑄)
𝑄̃ ∗𝐺 𝑄̃ ∗𝐺 𝑅

⎡ ⎛ ⎞⎤ Denote 𝜋𝑀 as the manufacturer’s profit. As 𝜋𝑀 = 𝜋𝐺 − 𝜋𝑅 , we have


𝑥𝐿
𝐺
⏞⏞⏞⏞
⏞ ⏞⏞⏞
⏞ ∞⎛ ⏞⏞⏞⏞⏞⏞⏞ ⎞
+ 𝑠1 ⎢𝑄 − ⎜𝑆 𝑈 (𝑥) + 𝑥 ⎟⎥ 𝑑𝐹 (𝑥) − 𝑚 ⎜𝑥 + 𝑆 𝐿 (𝑥) −𝑄⎟ 𝑑𝐹 (𝑥) (8)
∫0 ⎢ ⎜ 𝐺 ⎟⎥ 1∫
𝑈 ⎜ 𝐺 ⎟ 𝜋𝑀 =(𝑤0 − 𝑚0 )𝑄 + (𝑤1 − 𝑚1 )(𝑄̃ − 𝑄)+
𝑥𝐺
⎣ ⎝ ⎠⎦ ⎝ ⎠
− (𝑏1 − 𝑠1 )(𝑄̃ − 𝑄)− − (𝑏2 − 𝑠2 )𝐸𝑋 {𝐸𝑌 {(𝑄̃ − 𝑋 − 𝑌 )+ |𝑋 = 𝑥}}.
From the expected profit function above, 𝑄∗𝐺 , the optimal produc- (15)
tion quantity of 𝑄 can be determined by maximizing the expected
channel profit. With the wholesale price contract under the decentralized system,
the retailer and the manufacturer tend to optimize their own profits,
Proposition 4.1. The optimal channel production quantity 𝑄∗𝐺 is a and thus double marginalization happens. As such, the channel profit
solution of the equation: might not be maximized. The manufacturer and the retailer usually
coordinate the supply chain to solve this problem such that both of the
[𝑝2 − 𝑠1 ]𝐹 (𝑥𝐿 𝑈
𝐺 (𝑄)) − [𝑝2 − 𝑚1 ]𝐹 (𝑥𝐺 (𝑄)), manufacturer and the retailer gain higher profits and achieve Pareto
𝑥𝑈
𝐺
(𝑄) improvement. In the following section, we focus on how to set the
+ (𝑝2 − 𝑠2 ) 𝐻(𝑄 − 𝑥|𝑥)𝑑𝐹 (𝑥) = 𝑚1 − 𝑚0 . (9) values of 𝑤0 , 𝑤1 , 𝑏1 and 𝑏2 to coordinate the supply chain and allocate
∫𝑥𝐿 (𝑄)
𝐺 the channel profit.
We can see that the centralized decision on 𝑄 can maximize the
expected channel profit. 4.3. Efficient contract price for SCC
Because of the adjustment at the beginning of the second period, the
sale quantity is closer to the actual demand. Therefore, the centralized To realize SCC, we not only determine the order quantity at the
system yields the higher channel profit. beginning of the first period, but also adjust the order quantity based on
Next, we show how the manufacturer and the retailer share the the updated demand information at the beginning of the second period.
channel profit by coordinating the supply chain. The ideal situation In our setting, the retailer’s optimal order quantity should be equal
is that the optimal order quantity based on the coordination strategy to the channel’s optimal production quantity. Furthermore, we let the
(decentralized system) is the same as the optimal production quantity optimal values of 𝑆𝑅𝐿 (𝑆𝑅𝑈 ) be the same as 𝑆𝐺𝐿 (𝑆𝐺𝑈 ). Consequently, 𝑥𝐿
𝑅
=
𝑥𝐿 and 𝑈 = 𝑥𝑈 . We determine the proper values of the parameters (𝑤 ,
based on the centralized system. 𝐺
𝑥𝑅 𝐺 0
𝑤1 , 𝑏1 and 𝑏2 ) to coordinate the supply chain.
4.2. Decentralized system 𝑥𝑅 = 𝑥𝐺 and 𝑥𝑈
𝐿 𝐿
𝑅
= 𝑥𝑈
𝐺
can derive
𝑝2 − 𝑚1 𝑝 − 𝑤1
In the decentralized system, the manufacturer and the retailer com- = 2 , (16)
𝑝2 − 𝑠2 𝑝 2 − 𝑏2
mit an agreement on the transfer prices: 𝑤0 , 𝑤1 , 𝑏1 and 𝑏2 . Below 𝑝2 − 𝑠1 𝑝 − 𝑏1
we determine these parameters to maximize the channel profit and = 2 . (17)
𝑝2 − 𝑠2 𝑝2 − 𝑏2
to allocate the channel profit between the manufacturer and the re-
tailer, to thus realize the supply chain coordination. Under the given The retailer’s optimal order quantity should be equal to the chan-
pricing scheme, the retailer determines the optimal order quantity by nel’s optimal production quantity, i.e., 𝑄∗𝑅 = 𝑄∗𝐺 .
maximizing its own profit. With the above constraints in the decentralized system, the retailer
Similar to the channel profit function, the expected channel profit can determine the optimal order quantity at the beginning of the
function of the retailer 𝜋𝑅 (𝑄) is expressed as first period and the order adjustment at the beginning of the second
period, and thus the supply chain is coordinated. We have the following
̃
𝜋𝑅 = 𝑝1 𝐸𝑋 {𝑋} − 𝑤0 𝑄 + 𝐸𝑋 {max 𝑈𝑅 (𝑄|𝑋 = 𝑥)}, (10) proposition.
𝑄̃

5
X. Li et al. International Journal of Production Economics xxx (xxxx) xxx

Proposition 4.3. Suppose the wholesale price 𝑤0 is given. The supply The quantity of the reservation, 𝑋 follow a normal distribution
chain can be coordinated if with expectation 𝜇𝑥 and standard deviation 𝜎𝑥 . We assume that the
(𝑝2 − 𝑠2 )(𝑝2 − 𝑤0 ) demand in the sales period attracted by each reservation is independent
𝑏2 = 𝑝 2 − , (18) and identically normally distributed with parameters (𝜃, 𝛿). Since the
𝑝 2 − 𝑚0
(𝑝 − 𝑠1 )(𝑝2 − 𝑤0 ) central limit theorem provides an explanation of why the sum of
𝑏1 = 𝑝 2 − 2 , (19) independent normal random variables is normally distributed (Dunbar,
𝑝 2 − 𝑚0
2009), when the reservation quantity is 𝑋 = 𝑥, the total demand in the
(𝑝 − 𝑚1 )(𝑝2 − 𝑤0 )
𝑤1 = 𝑝 2 − 2 (20) sales period,√𝑌 is normally distributed with expectation 𝑥𝜃 and standard
𝑝2 − 𝑚0
deviation 𝛿 𝑥. Given the reservation quantity 𝑋 = 𝑥, the PDF of 𝑌 at
The value of the wholesale price 𝑤0 is in [𝑚0 , 𝑝2 ]. Essentially, if 𝑤0 = 𝑚0 , the beginning of the sales period is
then 𝑏2 = 𝑠2 , 𝑏1 = 𝑠1 , 𝑤1 = 𝑚1 , 𝜋𝑅 = 𝜋𝑇 and 𝜋𝑀 = 0; if 𝑤0 = 𝑝2 , then 2
1 − (𝑦−𝜃𝑥)
𝑏1 = 𝑏2 = 𝑤1 = 𝑝2 , 𝜋𝑅 = 0 and 𝜋𝑀 = 𝜋𝑇 . ℎ(𝑦|𝑥) = √ 𝑒 2𝑥𝛿2 , 𝑦 ≥ 0. (26)
𝛿 2𝑥𝜋
When the equations in Proposition 4.3 are satisfied, the supply chain
We analyze the impact of demand uncertainty on the order quantity,
can be coordinated. The value of 𝑤0 determines the channel profit
the thresholds of the quantity adjustment, and the expected channel
allocation between the manufacturer and the retailer.
profit. As the model is quite complicated, we conduct several numerical
experiments and illustrate the representative results.
4.4. Benchmarking model: Buy-back contract
With SCC (𝑄 = 𝑄∗𝑅 = 𝑄∗𝐺 , 𝑥𝐿 𝑅
= 𝑥𝐿𝐺
and 𝑥𝑈𝑅
= 𝑥𝑈 𝐺
), we adjust
the order quantity at the beginning of the second period according
We study a buy-back contract with supply chain coordination as a to Lemma 1. The results shown in Figs. 3(a) and 3(b) illustrate three
benchmarking model. In this buy-back contract, we assume that the points. (1) Both adjustment thresholds (𝑆𝐺𝑈 (𝑥)+𝑥 and 𝑆𝐺𝐿 (𝑥)+𝑥) increase
retailer is not allowed to adjust the order quantity at the beginning of with the reservation quantity (𝑥). (2) The value of 𝑥𝑈 − 𝑥𝐿 increases
𝐺 𝐺
the second period once the order is made at the beginning of the first with 𝛿, i.e., when the demand uncertainty is higher, we do not have to
period. The retailer returns the unsold items to the manufacturer at the adjust the order quantity. (3) Interestingly, if the reservation quantity
end of the second period and gets back a salvage value of 𝑏2 per item. is low, 𝑆𝐺𝑈 (𝑥) + 𝑥 will always increase in the uncertainty of the demand
The expected profit functions of the manufacturer, the retailer and the (𝛿). If the reservation quantity is high (e.g., 𝑥 = 100), as shown in
channel are as follows: Fig. 3(c), 𝑆𝐺𝐿 (𝑥) + 𝑥 will increase in the uncertainty of the demand (𝛿)
∞ 𝑄−𝑥
𝑏 when 𝑚1 < 6.5, but decrease in the uncertainty of the demand (𝛿) when
𝜋𝑀 =(𝑤0 − 𝑚0 )𝑄 + (𝑠2 − 𝑏2 ) (𝑄 − 𝑥 − 𝑦) 𝑑𝐻(𝑦|𝑥)𝑑𝐹 (𝑥), (21)
∫0 ∫0 𝑚1 > 6.5. The reason is that when the reservation quantity is high, as the
∞[ 𝑄−𝑥 ∞ ]
demand uncertainty increases, the high margin profit and low salvage
𝑏
𝜋𝑅 =𝑝2 𝑦𝑑𝐻(𝑦|𝑥) + (𝑄 − 𝑥)𝑑𝐻(𝑦|𝑥) 𝑑𝐹 (𝑥) loss (caused by a low manufacturing cost) motivate the supply chain
∫0 ∫0 ∫𝑄−𝑥
∞ to produce more goods, while the low margin profit and high salvage
+ 𝑝1 𝑥𝑑𝐹 (𝑥) loss (caused by a high manufacturing cost) prevent the supply chain
∫0
∞ 𝑄−𝑥 from producing more goods. Consequently, as shown in Fig. 3(b), if
+ 𝑏2 (𝑄 − 𝑥 − 𝑦) 𝑑𝐻(𝑦|𝑥)𝑑𝐹 (𝑥) − 𝑤0 𝑄, (22) the reservation quantity is high (e.g. 𝑥 ≥ 90), 𝑆𝐺𝐿 (𝑥) + 𝑥 will increase
∫0 ∫0
∞[ ]
𝑄−𝑥 ∞
in the uncertainty of the demand (𝛿) when 𝑚1 = 6, but decrease in the
𝑏
𝜋𝐺 =𝑝2 𝑦𝑑𝐻(𝑦|𝑥) + (𝑄 − 𝑥)𝑑𝐻(𝑦|𝑥) 𝑑𝐹 (𝑥) uncertainty of the demand (𝛿) when 𝑚1 = 8.
∫0 ∫0 ∫𝑄−𝑥

+ 𝑝1 𝑥𝑑𝐹 (𝑥) 5.1. Numerical study and model comparison
∫0
∞ 𝑄−𝑥
In the QACDL, the manufacturer and the retailer commit to an
+ 𝑠2 (𝑄 − 𝑥 − 𝑦) 𝑑𝐻(𝑦|𝑥)𝑑𝐹 (𝑥) − 𝑚0 𝑄. (23)
∫0 ∫0 agreement on the values of 𝑤0 , 𝑤1 , 𝑏1 and 𝑏2 , and the retailer decides
By maximizing 𝜋𝐺 𝑏 in (23), the channel optimal order quantity 𝑄𝑏 , the order quantity at the beginning of the first period. To illustrate the
𝐺
can be obtained by solving managerial insights, we set the parameters as follows: 𝑝1 = 9.5, 𝑝2 = 9.5,

𝑠1 = $4, 𝑠2 = 3, 𝑚0 = $5, 𝑚1 = $6, 𝜃 = 10, 𝛿 = {1, 2, 3, 4, 5}, 𝜇𝑥 = 75,
𝑝2 − 𝑚0 𝜎𝑥 = 10.
𝐻(𝑄 − 𝑥|𝑥)𝑑𝐹 (𝑥) = . (24)
∫0 𝑝2 − 𝑠2 We compare the performance of the supply chain between the two
Based on Cachon (2003), to coordinate the supply chain, (𝑏2 , 𝑤0 ) models (the QACDL model and the benchmarking model) to analyze the
should satisfy: impact of QA on the optimal order quantity and the two parties’ profits.
(𝑝2 − 𝑤0 )(𝑝2 − 𝑠2 ) We also analyze the impact of demand uncertainty on the supply chain
𝑏2 = 𝑝2 − . (25) performance improvement by analyzing the impact of factor 𝛿 on the
𝑝 2 − 𝑚0
optimal decisions and the corresponding channel profits.
Using this model as a benchmark, we analyze the improvement of As shown in Lemma 1, QA is directly affected by the two thresholds
the supply chain profit through the adjustment of order quantity for of the reservation quantity in the first period. When the reservation
different 𝛿. quantity is less than 𝑥𝐿 , the total supply is made up to 𝑆𝐺𝑈 (𝑥) + 𝑥,
𝐺
𝑈
where 𝑆𝐺 (𝑥) is the supply for the second period and 𝑥 is the supply for
5. Sensitive analysis and managerial insights the first period. When the reservation quantity is greater than 𝑥𝑈 𝐺
, the
total supply is reduced to 𝑆𝐺𝐿 (𝑥) + 𝑥. The two thresholds are affected by
In this section, we first assume that 𝑋 and 𝑌 are both normally decision variables 𝑄∗𝐺 , 𝑤1 , 𝑏1 , 𝑏2 , 𝑥𝐿
𝐺
and 𝑥𝑈
𝐺
. We summarize all of the
distributed in the numerical study. We test another distribution, say optimal decision variables in our model and those of the benchmarking
uniform distributions of 𝑋 and 𝑌 . We obtain similar results in the model (𝑏2 and 𝑄𝑏𝐺 ) in Table 1. Clearly, the order quantity 𝑄∗𝐺 is always
two schemes. As normally distributed demand is usual in reality and higher than 𝑄𝑏𝐺 , because the retailer has a chance to increase the order
in the literature, we provide the results with normally distributed amount at the beginning of the second period in the main model, but
demand in the text and the results with uniformly distributed demand not in the benchmarking model. As a result, the retailer orders more in
in Appendix B. the benchmarking model to avoid shortage.

6
X. Li et al. International Journal of Production Economics xxx (xxxx) xxx

Fig. 3. Illustration of the effects of 𝑚1 and 𝛿 on 𝑄̃ ∗𝐺 (𝑝1 = 10, 𝑝2 =10, 𝑠1 =4, 𝑠2 = 3, 𝑚0 =5, 𝜇𝑥 =75, 𝜎𝑥 =10, 𝜃=10, 𝛿={5,10,15}).

Table 1 QA, the performance of the main model is better than that of the
Generation of optimal decision variables under different 𝛿 with normally distributed
benchmarking model for different 𝛿. For example, when 𝛿 = 1, in the
demand. (𝑝1 = 9.5, 𝑝2 = 9.5, 𝑠1 = $4, 𝑠2 = $3, 𝑚0 = $5, 𝑚1 = $6, 𝜃 = 10, 𝜇𝑥 = 75, 𝜎𝑥 = 10).
benchmarking model, the retailer’s profit is $2307 and the manufac-
The optimal variables 𝛿=1 𝛿=2 𝛿=3 𝛿=4 𝛿=5
turer’s profit is $1153. In the main model, the retailer’s profit is $2405
Benchmarking model 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8 and the manufacturer’s profit is $1202. The profit of each party in the
𝑏2 $7.33 $7.33 $7.33 $7.33 $7.33
𝑄𝑏𝐺 880 881 882 883 884
main model is higher than that in the benchmarking model.
In the benchmarking model, we use a buy-back contract (Cachon,
Main model 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8
𝑤1 $8.3333 $8.3333 $8.3333 $8.3333 $8.3333
2003) to realize SCC. Due to QA, however, the optimal expected
𝑏1 $7.6667 $7.6667 $7.6667 $7.6667 $7.6667 channel profit in the main model is higher than the profit in the
𝑏2 $7.3333 $7.3333 $7.3333 $7.3333 $7.3333 benchmarking model.
𝑥𝐿𝐺 74.6 74.2 73.8 73.4 73.1 Therefore, in the main model, the supply chain takes advantages of
𝑥𝑈𝐺 75.3 75.7 76.0 76.3 76.7
both QA and SCC: 1. under SCC, the decentralized decisions are the
𝑄∗𝐺 829 834 838 843 847
same as the centralized decisions; 2. using the reservation strategy, the
manufacturer can have a longer lead time and reduce the production
cost, and the reservation information provides a more precise estimate
Table 2
of the demand; 3. with QA at the beginning of the second period, the
Generation of optimal expected profit under different 𝛿 with normally distributed
demand. (𝑝1 = 9.5, 𝑝2 = 9.5, 𝑠1 = $4, 𝑠2 = $3, 𝑚0 = $5, 𝑚1 = $6, 𝜃 = 10, 𝜇𝑥 = 75, supply can match the demand better than in the benchmarking model.
𝜎𝑥 = 10). These above advantages benefit both the manufacturer and the retailer
The optimal profit 𝛿=1 𝛿=2 𝛿=3 𝛿=4 𝛿=5 and thus achieves Pareto improvement.
Benchmarking model Using the same parameters and assumptions as in Table 1, we
(21) and (24): 𝜋𝑀𝑏
$2307 $2305 $2302 $2299 $2294 illustrate the application of Proposition 4.3. Table 2 shows, when 𝛿 = 1,
(22) and (24): 𝜋𝑅𝑏 $1153 $1152 $1151 $1149 $1147 the calculated optimal channel profit is $2405. Fig. 4 shows that the
(23) and (24): 𝜋𝐺𝑏 $3460 $3457 $3453 $3448 $3441 sum of the optimal manufacturer profit and the optimal retailer profit
Main model is always equal to the optimal channel profit under the QACDL contract.
(5), (9) and (10): 𝜋𝑀 $2405 $2393 $2381 $2368 $2356 Moreover, the wholesale price 𝑤0 determines the allocation of the
(5), (9) and (15): 𝜋𝑅 $1202 $1196 $1190 $1184 $1178
(5), (9) and (8): 𝜋𝐺 $3607 $3589 $3571 $3552 $3534
channel profit between the manufacturer and the retailer.

5.2. Impact of margin profit on order quantity and channel profit

The optimal channel profits in the benchmarking model and the In this subsection, we analyze the impact of 𝛿 on the relative
main model are summarized in Table 2. The results show that with difference of the order quantity 100(𝑄∗𝐺 − 𝑄𝑏𝐺 )∕𝑄𝑏𝐺 and the relative

7
X. Li et al. International Journal of Production Economics xxx (xxxx) xxx

Acknowledgments

This research was funded in part by the Grant of the Univer-


sity of Macau, Macao SAR, China (File no. MYRG2016-00016-FBA
and MYRG2019-00031-FBA), the Science and Technology Develop-
ment Fund, Macao SAR, China (File no. FDCT/027/2016/A1), and the
National Natural Science Foundation of China under the Grant no.
71801233 and the Grant no. 71701182.

Appendix A

Fig. 4. The allocation of the channel profit at different 𝑤0 (𝑝1 = 9.5, 𝑝2 = 9.5, 𝑠1 = $4,
Proof of Lemma 1. Based on the reservation quantity 𝑥, the updated
𝑠2 = $3, 𝑚0 = $5, 𝑚1 = $6, 𝜃 = 10, 𝛿 = 1, 𝜇𝑥 =75, 𝜎𝑥 =10).
demand distribution during the second period can be obtained and the
production quantity will be adjusted to 𝑄̃ at the beginning of the second
𝑏 )∕𝜋 𝑏 . As both the order
period. Define 𝑆 as the total supply during the second period. Then
difference of the channel profit 100(𝜋𝐺 − 𝜋𝐺 𝐺 𝑄̃ = 𝑥+𝑆. In this Lemma, we will develop an optimal production policy
quantity and the thresholds that determine QA are not affected by 𝑝1 to adjust the production quantity from 𝑄 to 𝑄. ̃
but by 𝑝2 , by comparing the results of different values of retail price Let
𝑝2 , we show the joint impacts of 𝛿 and 𝑝2 on the relative order quantity
difference and the relative channel profit difference, respectively. 𝐺𝑐 (𝑆) = −𝑠1 𝑆 + 𝐺(𝑆), (27)
As shown in Fig. 5(a), 100(𝑄∗𝐺 − 𝑄𝑏𝐺 )∕𝑄𝑏𝐺 is negative under the 𝐺𝑐 (𝑆) = −𝑚1 𝑆 + 𝐺(𝑆), (28)
condition of 𝑝2 = 9.5 but positive under the condition of 𝑝2 = 6.5. In
where
other words, with QA at the beginning of the second period, as shown
in Fig. 5(c), the retailer orders more low-margin-profit products but less 𝐺(𝑆) = 𝑝2 𝐸𝑌 {min{𝑆, 𝑌 }} + 𝑠2 𝐸𝑌 {𝑆 − 𝑌 }+ . (29)
high-margin-profit products at the beginning of the first period. As the
order quantity does not need to be adjusted when the demand uncer- Taking the first order derivative of (27) and (28) with respect to 𝑆, we
obtain
tainty is high enough, 100(𝑄∗𝐺 − 𝑄𝑏𝐺 )∕𝑄𝑏𝐺 approaches to 0. Therefore, ( )
100(𝑄∗𝐺 − 𝑄𝑏𝐺 )∕𝑄𝑏𝐺 increases in 𝛿 when the product’s margin profit is 𝑝2 − 𝑚1
𝑆𝐺𝐿 (𝑥) = 𝐻 −1 |𝑥 , (30)
high (caused by a high price), but decreases in 𝛿 when the product’s 𝑝 − 𝑠2
( 2 )
margin profit is low (caused by a low price). As the profit is higher with 𝑝2 − 𝑠1
𝑆𝐺𝑈 (𝑥) = 𝐻 −1 |𝑥 . (31)
an order quantity adjustment than without an adjustment, as shown 𝑝2 − 𝑠2
in Fig. 5(b), 100(𝜋𝐺 ∗ − 𝜋 𝑏 )∕𝜋 𝑏 is always positive. When the demand
𝐺 𝐺 Let
uncertainty is high enough, the order need not be adjusted even in
the main model. As a result, 100(𝜋𝐺 ∗ − 𝜋 𝑏 )∕𝜋 𝑏 always decreases in 𝛿 𝑀(𝑥) = 𝑠1 (𝑄 − 𝑥) + max {−𝑠1 𝑆 + 𝐺(𝑆)}, (32)
𝐺 𝐺 𝑆+𝑥<𝑄
and converges to 0. As shown in Fig. 5(d), 100(𝜋𝐺 ∗ − 𝜋 𝑏 )∕𝜋 𝑏 increases
𝐺 𝐺 𝑁(𝑥) = 𝑚1 (𝑄 − 𝑥) + max {−𝑚1 𝑆 + 𝐺(𝑆)}, (33)
and then decreases in 𝑝2 . In other words, with QA, the supply chain 𝑆+𝑥>𝑄

performance improvement for moderate-margin-profit goods is better Next, we will consider three cases: 𝑄 ≤ 𝑆𝐺𝐿 + 𝑥, 𝑆𝐺𝐿 + 𝑥 < 𝑄 ≤ 𝑆𝐺𝑈 + 𝑥
than the performance improvement for high-margin-profit goods or and 𝑆𝐺𝑈 + 𝑥 < 𝑄.
low-margin-profit goods. If 𝑄 ≤ 𝑆𝐺𝐿 + 𝑥, then 𝑄 < 𝑆𝐺𝑈 + 𝑥,

𝑁(𝑥) =𝑚1 (𝑄 − 𝑥) − 𝑚1 𝑆𝐺𝐿 + 𝐺(𝑆𝐺𝐿 )


6. Conclusion
≥𝑚1 (𝑄 − 𝑥) − 𝑚1 (𝑄 − 𝑥) + 𝐺(𝑄 − 𝑥)

Increasing profit is crucial for a company. In this paper, we com- =𝐺(𝑄 − 𝑥) = 𝑀(𝑥). (34)
bine the reservation marketing strategy with the two-mode production Consequently, 𝑆 ∗ = 𝑆𝐺𝐿 and 𝑄̃ ∗𝐺 = 𝑆𝐺𝐿 + 𝑥.
strategy to increase the channel profit and generate a new contract. If 𝑆𝐺𝐿 + 𝑥 < 𝑄 ≤ 𝑆𝐺𝑈 + 𝑥, then
The contract that we generate is a suitable tool for the two-period
case in which one period is the reservation period and the other is the 𝑁(𝑥) = 𝐺(𝑄 − 𝑥) = 𝑀(𝑥).
sales period. Through the contract, the profits of both parties in the Consequently, 𝑆 ∗ = 𝑄 − 𝑥 and 𝑄̃ ∗𝐺 = 𝑄.
supply chain are increased through QA (providing better matching of If 𝑆𝐺𝑈 + 𝑥 < 𝑄, then 𝑆𝐺𝐿 + 𝑥 < 𝑄,
supply with demand) and SCC (generating the supply chain’s optimal
solution). 𝑀(𝑥) =𝑠1 (𝑄 − 𝑥) − 𝑠1 𝑆𝐺𝑈 + 𝐺(𝑆𝐺𝑈 )
We show the impact of demand uncertainty on the order quantity, ≥𝑠1 (𝑄 − 𝑥) − 𝑠1 (𝑄 − 𝑥) + 𝐺(𝑄 − 𝑥)
the thresholds of QA and the expected channel profit. The updated =𝐺(𝑄 − 𝑥) = 𝑁(𝑥). (35)
information plays an essential role in the ordering policy. Our model
can be applied in the case of general demand distribution. The model Consequently, 𝑆 ∗ = 𝑆𝐺𝑈 and 𝑄̃ ∗𝐺 = 𝑆𝐺𝑈 + 𝑥.
can assist companies in making ordering policy and adjusting the However 𝑆𝐺𝐿 and 𝑆𝐺𝑈 are both a function of 𝑥. Let 𝑥𝑈
𝐺
solves
order quantity in two-period case. We analyze the impact of demand 𝑄 = 𝑆𝐺𝐿 (𝑥) + 𝑥,
uncertainty on supply chain performance improvement through QA.
We show that when demand uncertainty is high enough, the order and 𝑥𝐿
𝐺
solves
need not be adjusted even with the contract in the main model. As 𝑄 = 𝑆𝐺𝑈 (𝑥) + 𝑥.
a result, the increment/decrement part of the order quantity due to
QA decreases in demand uncertainty, such that the improvement of the Consequently, the three cases are equivalent to 𝑥 ≥ 𝑥𝑈
𝐺
, 𝑥𝐿
𝐺
≤ 𝑥 < 𝑥𝑈
𝐺
𝐿
and 𝑥 < 𝑥𝐺 . Then, we can obtain the result. □
expected channel profit also decreases with demand uncertainty.

8
X. Li et al. International Journal of Production Economics xxx (xxxx) xxx

Fig. 5. Illustration of the effect of 𝛿 and 𝑝2 on 100(𝑄∗𝐺 − 𝑄𝑏𝐺 )∕𝑄𝑏𝐺 and 100(𝜋𝐺 − 𝜋𝐺𝑏 )∕𝜋𝐺𝑏 (𝑝1 = 6.5, 𝑝2 ={6.5,9.5}, 𝑠1 =4, 𝑠2 = 3, 𝑚0 =5, 𝑚1 =6, 𝑥𝑈 =100, 𝑥𝐿 =50, 𝜃 = 10, 𝑑=[0,1]).

(𝑝2 − 𝑚1 )(𝑝2 − 𝑤0 )
Proof of Proposition 4.1. Based on (8), taking the first and the second 𝑤1 = 𝑝 2 − (38)
order derivative of 𝜋𝐺 , we obtain 𝑝2 − 𝑚0

𝜕𝜋𝐺 Coupled with (10) and (15), we can derive the value of the whole-
=𝑚1 − 𝑚0 − [𝑝2 − 𝑠1 ]𝐹 (𝑥𝐿 𝑈
𝐺 (𝑄)) + [𝑝2 − 𝑚1 ]𝐹 (𝑥𝐺 (𝑄)), sale price 𝑤0 is in [𝑚0 , 𝑝2 ]. Essentially, if 𝑤0 = 𝑚0 , 𝜋𝑅 = 𝜋𝑇 while
𝜕𝑄
𝑥𝑈 (𝑄) 𝜋𝑀 = 0; if 𝑤0 = 𝑝2 , 𝜋𝑅 = 0 while 𝜋𝑀 = 𝜋𝑇 . □
𝐺
− (𝑝2 − 𝑠2 ) 𝐻(𝑥, 𝑄 − 𝑥)𝑑𝐹 (𝑥).
∫𝑥𝐿 (𝑄)
𝐺 Appendix B. Uniform reservation distribution and uniform condi-
𝜕 2 𝜋𝐺 𝑥𝑈 (𝑄)
𝐺 tional demand distribution
= − (𝑝2 − 𝑠2 ) ℎ(𝑥, 𝑄 − 𝑥)𝑑𝐹 (𝑥) < 0.
(𝜕𝑄)2 ∫𝑥𝐿 (𝑄)
𝐺
The distribution of 𝑋 is uniformed with the upper bound of 𝑥𝑈 and
Making 𝜕𝜋𝐺 ∕𝜕𝑄 = 0, we obtain the result. □
the lower bound of 𝑥𝐿 . The CDF of 𝑥 is
𝐿 [ ]
Proof of Proposition 4.2. The proof is similar with Proposition 4.1. □ 𝐹 (𝑥) = 𝑥𝑥−𝑥
𝑈 −𝑥𝐿 , 𝑥 ∈ 𝑥𝐿 , 𝑥𝑈 . (39)

Proof of Proposition 4.3. Combined with (16) and (17), we have We assume that the demand quantity depends on the reservation
quantity. In order to illustrate the impact of the reservation quantity
(𝑝2 − 𝑏2 )(𝑝2 − 𝑠1 )
𝑏1 =𝑝2 − , on demand, we set a certain distribution for both the reservation and
(𝑝2 − 𝑠2 )
demand, uniform distribution. When the reservation quantity in the
(𝑝 − 𝑏2 )(𝑝2 − 𝑚1 )
𝑤1 =𝑝2 − 2 . beginning of the second period 𝑋 = 𝑥, the demand distribution in sale
(𝑝2 − 𝑠2 )
season, 𝑌 is assumed to be uniformed in [(1 − 𝑑)𝜃𝑥, (1 + 𝑑)𝜃𝑥], where 𝜃
Besides, let 𝑄∗𝑅 = 𝑄∗𝐺 . Coupled with (14), we derive is the mean sensitivity to the reservation based demand and 𝑑 ∈ [0, 1]
(𝑝2 − 𝑠2 )(𝑝2 − 𝑤0 ) is the fluctuation factor relative to 𝜃. Then we have
𝑏2 =𝑝2 − , 𝑦 − (1 − 𝑑)𝜃𝑥
𝑇 𝐻(𝑦|𝑥) = , 𝑦 ∈ [(1 − 𝑑)𝜃𝑥, (1 + 𝑑)𝜃𝑥] , (40)
(𝑝 − 𝑠1 )(𝑝2 − 𝑤0 ) 2𝑑𝜃𝑥
𝑏1 =𝑝2 − 2
𝑇 Then we can obtain the CDF of demand as follows:
(𝑝2 − 𝑚1 )(𝑝2 − 𝑤0 )
𝑤1 =𝑝2 − ,
𝑇 Lemma 2. If (1 + 𝑑)𝑥𝐿 ≤ (1 − 𝑑)𝑥𝑈 , then
[
] 𝑥𝑈 (
where 𝑇 = (𝑝2 − 𝑠1 )𝐹 (𝑥𝐿
𝐺
) + (𝑝2 − 𝑚1 ) 1 − 𝐹 (𝑥𝑈
𝐺
) + (𝑝2 − 𝑠2 ) ∫ 𝐿𝐺 𝐻 𝑄∗𝐺 ⎧
𝑥𝐺
⎪ 𝑦[ln 𝑦−1−ln (1−𝑑)𝜃𝑥𝐿 ]+(1−𝑑)𝜃𝑥𝐿
−𝑥|𝑥) 𝑑𝐹 (𝑥). 𝑥𝐿
𝐺
, 𝑥𝑈
𝐺
and 𝑄∗𝐺 can be obtain in (6), (7) and (9). ⎪ 2𝑑𝜃(𝑥𝑈 −𝑥𝐿 )
, (1 − 𝑑)𝜃𝑥𝐿 ≤ 𝑦 < (1 + 𝑑)𝜃𝑥𝐿 ;
Combined with (9), we obtain ⎪ 1+𝑑
𝑦 ln 1−𝑑 −2𝑑𝜃𝑥𝐿
𝐻(𝑦) = ⎨ , (1 + 𝑑)𝜃𝑥𝐿 ≤ 𝑦 < (1 − 𝑑)𝜃𝑥𝑈 ;
(𝑝 − 𝑠2 )(𝑝2 − 𝑤0 ) ⎪ 2𝑑𝜃(𝑥𝑈 −𝑥𝐿 )
𝑏2 = 𝑝2 − 2 , (36) ⎪ 𝑦[1+ln 𝑥𝑈 𝜃(1+𝑑)−ln 𝑦]−2𝑑𝜃𝑥𝐿 −(1−𝑑)𝜃𝑥𝑈
𝑝 2 − 𝑚0 ⎪ , (1 − 𝑑)𝜃𝑥𝑈 ≤ 𝑦 ≤ (1 + 𝑑)𝜃𝑥𝑈 .
2𝑑𝜃(𝑥𝑈 −𝑥𝐿 )
(𝑝 − 𝑠1 )(𝑝2 − 𝑤0 ) ⎩
𝑏1 = 𝑝 2 − 2 , (37)
𝑝 2 − 𝑚0 (41)

9
X. Li et al. International Journal of Production Economics xxx (xxxx) xxx

While if (1 + 𝑑)𝑥𝐿 > (1 − 𝑑)𝑥𝑈 , then


⎪ 𝑦{ln 𝑦−1−ln[(1−𝑑)𝜃𝑥𝐿 ]}+(1−𝑑)𝜃𝑥𝐿
, (1 − 𝑑)𝜃𝑥𝐿 ≤ 𝑦 < (1 − 𝑑)𝜃𝑥𝑈 ;
⎪ 2𝑑𝜃(𝑥𝑈 −𝑥𝐿 )
⎪ 𝑦 ln 𝑥𝑈
−(1−𝑑)𝜃(𝑥𝑈 −𝑥𝐿 )
𝐻(𝑦) = ⎨ 𝑥𝐿
, (1 − 𝑑)𝜃𝑥𝑈 ≤ 𝑦 < (1 + 𝑑)𝜃𝑥𝐿 ;
⎪ 2𝑑𝜃(𝑥𝑈 −𝑥𝐿 )
⎪ 𝑦{1+ln[ 𝑥𝑈 𝜃(1+𝑑) ]−ln 𝑦}−2𝑑𝜃𝑥𝐿 −(1−𝑑)𝜃𝑥𝑈
⎪ 2𝑑𝜃(𝑥𝑈 −𝑥𝐿 )
, (1 + 𝑑)𝜃𝑥𝐿 ≤ 𝑦 ≤ (1 + 𝑑)𝜃𝑥𝑈 .

(42)
Take the first order derivative of 𝐻(𝑦) with respect to 𝑦, the PDF of
demand is as follows: If (1 + 𝑑)𝑥𝐿 ≤ (1 − 𝑑)𝑥𝑈 , then
𝑦
ln
⎧ (1−𝑑)𝜃𝑥𝐿
, (1 − 𝑑)𝜃𝑥𝐿 ≤ 𝑦 < (1 + 𝑑)𝜃𝑥𝐿 ;
⎪ 2𝑑𝜃(𝑥𝑈 −𝑥𝐿 )
⎪ ln 1+𝑑 Fig. 6. Illustration of the effects of 𝑑 on 𝑄̃ ∗𝐺 with uniform distributed demand (𝑥𝐿 =50,
ℎ(𝑦) = ⎨ 1−𝑑
, (1 + 𝑑)𝜃𝑥𝐿 ≤ 𝑦 < (1 − 𝑑)𝜃𝑥𝑈 ; (43)
2𝑑𝜃(𝑥𝑈 −𝑥𝐿 ) 𝑥𝑈 =100, 𝜃=10, 𝑑=[0.4,1], 𝑝=10, 𝑠1 =4, 𝑠2 = 3, 𝑚0 =5, 𝑚1 =6).
⎪ 𝑈
⎪ ln (1+𝑑)𝜃𝑥
𝑦
⎩ 2𝑑𝜃(𝑥𝑈 −𝑥𝐿 )
, (1 − 𝑑)𝜃𝑥𝑈 ≤ 𝑦 ≤ (1 + 𝑑)𝜃𝑥𝑈 .
Table 3
While if (1 + 𝑑)𝑥𝐿 > (1 − 𝑑)𝑥𝑈 , then Generation of optimal decision variables at different 𝑑 with uniform distributed
demand. (𝑝1 = 9.5, 𝑝2 = 9.5, 𝑠1 = $4, 𝑠2 = $3, 𝑚0 = $5, 𝑚1 = $6, 𝜃 = 10, 𝑥𝑈 = 100,
⎧ ln
𝑦 𝑥𝐿 = 50, 𝑑 = [0.1, 0.5]).
⎪ (1−𝑑)𝜃𝑥𝐿
, (1 − 𝑑)𝜃𝑥𝐿 ≤ 𝑦 < (1 − 𝑑)𝑥𝑈 ; The optimal variables 𝑑 = 0.1 𝑑 = 0.2 𝑑 = 0.3 𝑑 = 0.4 𝑑 = 0.5
⎪ 2𝑑𝜃(𝑥 −𝑥𝐿 )
𝑈
⎪ 𝑈
ln 𝑥 𝐿 Benchmarking model 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8
ℎ(𝑦) = ⎨ 𝑥
, (1 − 𝑑)𝜃𝑥𝑈 ≤ 𝑦 < (1 + 𝑑)𝜃𝑥𝐿 ; (44)
2𝑑𝜃(𝑥 −𝑥𝐿 )
𝑈 𝑏2 $7.33 $7.33 $7.33 $7.33 $7.33
⎪ (1+𝑑)𝜃𝑥𝑈 𝑄𝑏𝐺 821.2264 848.9705 876.7147 904.4588 932.2029
⎪ ln 𝑦
⎪ 2𝑑𝜃(𝑥𝑈 −𝑥𝐿 )
, (1 + 𝑑)𝜃𝑥𝐿 ≤ 𝑦 ≤ (1 + 𝑑)𝜃𝑥𝑈 . Main model 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8 𝑤0 = 8

𝑤1 $8.3333 $8.3333 $8.3333 $8.3333 $8.3333
𝑏1 $7.6667 $7.6667 $7.6667 $7.6667 $7.6667
Proof of Lemma 2. As 𝑏2 $7.3333 $7.3333 $7.3333 $7.3333 $7.3333
𝑦−(1−𝑑)𝜃𝑥 𝑥𝐿𝐺 73 71 70 68 67
𝐻(𝑦|𝑥) = 2𝑑𝜃𝑥
, 𝑦 ∈ [(1 − 𝑑)𝜃𝑥, (1 + 𝑑)𝜃𝑥] , 𝑥𝑈𝐺 77 79 81 83 85
𝑄∗𝐺 854 882 910 937 965
where the cumulative distribution function of 𝑅 is
𝑥 − 𝑥𝐿 [ ]
𝐹 (𝑥) = 𝑈 , 𝑦 ∈ 𝑥𝐿 , 𝑥𝑈 . Table 4
𝑥 − 𝑥𝐿
Generation of optimal expected profit at different 𝑑 with uniform distributed demand.
We obtain that (𝑝1 = 9.5, 𝑝2 = 9.5, 𝑠1 = $4, 𝑠2 = $3, 𝑚0 = $5, 𝑚1 = $6, 𝜃 = 10, 𝑥𝑈 = 100, 𝑥𝐿 = 50,
𝑥𝑈 𝑦 𝑑 = [0.1, 0.5]).
1
𝐻(𝑦) = ( ) 𝑑𝑧𝑑𝑥 The optimal profit 𝑑 = 0.1 𝑑 = 0.2 𝑑 = 0.3 𝑑 = 0.4 𝑑 = 0.5
∫𝑥𝐿 ∫(1−𝑑)𝜃𝑥 2𝑑𝜃 𝑥𝑈 − 𝑥𝐿 𝑥
𝑥𝑈 𝑦 Benchmarking model
∫𝑥𝐿 ∫(1−𝑑)𝜃𝑥 𝑥1 𝑑𝑧𝑑𝑥 (21) and (24): 𝜋𝑀𝑏
$2167 $2163 $2132 $2085 $2029
= ( ) ,𝑦 ∈ [(1 − 𝑑)𝜃𝑥, (1 + 𝑑)𝜃𝑥]. (45) (22) and (24): 𝜋𝑅𝑏 $1083 $1081 $1066 $1042 $1015
2𝑑𝜃 𝑥𝑈 − 𝑥𝐿
(23) and (24): 𝜋𝐺𝑏 $3250 $3244 $3199 $3127 $3044
We need to exchange the order of the integration. As the function Main model
above shown, (5), (9) and (10): 𝜋𝑀 $2319 $2253 $2188 $2121 $2055
𝐿 𝑈
(5), (9) and (15): 𝜋𝑅 $1159 $1126 $1094 $1061 $1027
𝑥 ≤𝑥≤𝑥 , (46) (5), (9) and (8): 𝜋𝐺 $3478 $3380 $3281 $3182 $3083
(1 − 𝑑)𝜃𝑥 ≤ 𝑦 ≤ (1 + 𝑑)𝜃𝑥. (47)
Given 𝑦, the domain of 𝑥 may not cover the full region of [𝑥𝐿 , 𝑥𝑈 ].
We first to assume that (1 + 𝑑)𝜃𝑥𝐿 < (1 − 𝑑)𝜃𝑥𝑈 , and the proof is similar Simply the function above, we can obtain the result.
for (1 + 𝑑)𝜃𝑥𝐿 < (1 − 𝑑)𝜃𝑥𝑈 . It is similar to obtain the CDF of demand for the case of (1+𝑑)𝜃𝑥𝐿 <
Based on (46) and (47), we obtain (1 − 𝑑)𝜃𝑥𝑈 . □
𝑦
𝑥𝐿 ≤ 𝑥 ≤ , if (1 − 𝑑)𝜃𝑥𝐿 ≤ 𝑦 < (1 + 𝑑)𝜃𝑥𝐿 , The results with uniformly distributed demand can be found in
𝜃(1 − 𝑑)
𝑦 𝑦 Fig. 6, Tables 3 and 4.
≤𝑥≤ , if (1 + 𝑑)𝜃𝑥𝐿 ≤ 𝑦 < (1 − 𝑑)𝜃𝑥𝑈 ,
𝜃(1 + 𝑑) 𝜃(1 − 𝑑)
𝑦 References
≤ 𝑥 ≤ 𝑥𝑈 , if (1 − 𝑑)𝜃𝑥𝑈 ≤ 𝑦 < (1 + 𝑑)𝜃𝑥𝑈 ,
𝜃(1 + 𝑑)
Consequently, the CDF of demand is Aviv, Y., Wei, M.M., Zhang, F., 2019. Responsive pricing of fashion products: the effects
of demand learning and strategic consumer behavior. Management Science.
⎧ 𝑦
𝑧
𝜃(1−𝑑) 1
Barnes-Schuster, D., Bassok, Y., Anupindi, R., 2002. Coordination and flexibility in
⎪ ∫ ∫
(1−𝑑)𝜃𝑥𝐿 𝑥𝐿 𝑥
𝑑𝑥𝑑𝑧 supply contracts with options. Manuf. Serv. Oper. Manag. 4 (3), 171–207.
⎪ 2𝑑𝜃(𝑥𝑈 −𝑥𝐿 )
, (1 − 𝑑)𝜃𝑥𝐿 ≤ 𝑦 < (1 + 𝑑)𝜃𝑥𝐿 ;
Cachon, G.P., 2003. Supply chain coordination with contracts. Handbooks Oper. Res.
⎪ 𝑧
(1+𝑑)𝜃𝑥𝐿
⎪ ∫ ∫
𝜃(1−𝑑) 1
𝑑𝑥𝑑𝑧+∫
𝑦 𝜃+𝑑
ln 𝜃−𝑑 𝑑𝑧 Management Sci. 11, 229–340.
(1−𝑑)𝜃𝑥𝐿 𝑥𝐿 𝑥 (1+𝑑)𝜃𝑥𝐿
⎪ 𝑈 𝐿 , Cachon, G.P., Lariviere, M.A., 2005. Supply chain coordination with revenue-sharing
⎪ 2𝑑𝜃(𝑥 −𝑥 )
contracts: Strengths and limitations. Manage. Sci. 51 (1), 30–44.
𝐻(𝑦) = ⎨
⎪ (1 + 𝑑)𝜃𝑥𝐿 ≤ 𝑦 < (1 − 𝑑)𝜃𝑥𝑈 ; Chen, X., Shum, S., Simchi-Levi, D., 2014. Stable and coordinating contracts for a supply
𝑧 chain with multiple risk-averse suppliers. Prod. Oper. Manage. 23 (3), 379–392.
⎪ ∫
(1+𝑑)𝜃𝑥𝐿

𝜃(1−𝑑) 1
𝑑𝑥𝑑𝑧+∫
(1−𝑑)𝜃𝑥𝑈 𝜃+𝑑
ln 𝜃−𝑑 𝑑𝑧+∫
𝑦
ln (1+𝑑)𝜃𝑥
𝑈
𝑑𝑧
⎪ (1−𝑑)𝜃𝑥𝐿 𝑥𝐿 𝑥 (1+𝑑)𝜃𝑥𝐿 (1−𝑑)𝜃𝑥𝑈 𝑧 Choi, T.M., Li, D., Yan, H., 2003. Optimal two-stage ordering policy with bayesian
⎪ 𝑈
2𝑑𝜃(𝑥 −𝑥 )𝐿 , information updating. J. Oper. Res. Soc. 54 (8), 846–859.
⎪ Dockner, E.J., Fruchter, G.E., 2014. Coordinating production and marketing with
⎪ (1 − 𝑑)𝜃𝑥𝑈 ≤𝑦≤ (1 + 𝑑)𝜃𝑥𝑈 . dynamic transfer prices. Prod. Oper. Manage. 23 (3), 431–445.

10
X. Li et al. International Journal of Production Economics xxx (xxxx) xxx

Dong, L., Zhu, K., 2007. Two-wholesale-price contracts: Push, pull, and advance- Modak, N.M., Panda, S., Sana, S.S., 2015. Three-echelon supply chain coordination
purchase discount contracts. Manuf. Serv. Oper. Manag. 9 (3), 291–311. considering duopolistic retailers with perfect quality products. Int. J. Prod. Econ.
Donohue, K.L., 2000. Efficient supply contracts for fashion goods with forecast updating Panda, S., Modak, N., Basu, M., Goyal, S., 2015. Channel coordination and profit
and two production modes. Manage. Sci. 46 (11), 1397–1411. distribution in a social responsible three-layer supply chain. Int. J. Prod. Econ.
Dunbar, S.R., 2009. Topics in probability theory and stochastic processes. Department 168, 224–233.
of Mathematics. University of Nebraska, Lincoln/NE. Pasternack, B.A., 2008. Optimal pricing and return policies for perishable commodities.
Dvoretzky, A., Kiefer, J., Wolfowitz, J., 1952. The inventory problem: Ii case of Mark. Sci. 27 (1), 133–140.
unknown distributions of demand. Econometrica 20 (3), 450–466. Saha, S., Goyal, S., 2015. Supply chain coordination contracts with inventory level and
Eppen, G.D., Iyer, A.V., 1997. Improved fashion buying with bayesian updates. Oper. retail price dependent demand. Int. J. Prod. Econ. 161, 140–152.
Res. 45 (6), 805–819. Sethi, S., Sorger, G., 1991. A theory of rolling horizon decision making. Ann. Oper.
Gallego, G., Özer, Ö., 2001. Integrating replenishment decisions with advance demand Res. 29 (1), 387–415.
information. Manage. Sci. 47 (10), 1344–1360. Sethi, S.P., Yan, H., Zhang, H., 2004. Quantity flexibility contracts: Optimal decisions
Ghemawat, P., 2003. Zara: fast fashion. Harvard Business School case 9-703-497. with information updates. Decision Sciences 35 (4), 691–712.
Gurnani, H., Tang, C.S., 1999. Note: Optimal ordering decisions with uncertain cost Sethi, S.P., Yan, H., Zhang, H., 2006. Inventory and Supply Chain Management with
and demand forecast updating. Manage. Sci. 45 (10), 1456–1462. Forecast Updates, Vol. 81. Springer.
Hausmann, W.H., 1969. Sequential decision problems: A model to exploit existing Shih, C.-C., Lin, T.M., Luarn, P., 2014. Fan-centric social media: The xiaomi
forecast. Manage. Sci. 16 (2), B93–B111. phenomenon in China. Bus. Horiz. 57 (3), 349–358.
Heath, D.C., Jackson, P.L., 1994. Modeling the evolution of demand forecasts ith Spengler, J.J., 1950. Vertical integration and antitrust policy. J. Polit. Econ. 58 (4),
application to safety stock analysis in production/distribution systems. IIE Trans. 347–352.
26 (3), 17–30. Tsay, A.A., Lovejoy, W.S., 1999. Quantity flexibility contracts and supply chain
Henig, M., Gerchak, Y., Ernst, R., Pyke, D.F., 1997. An inventory model embedded in performance. Manuf. Serv. Oper. Manag. 1 (2), 89–111.
designing a supply contract. Manage. Sci. 43 (2), 184–189. Wei, M.M., Zhang, F., 2018. Advance selling to strategic consumers: Preorder contingent
Hu, B., Meng, C., Xu, D., Son, Y.J., 2016. Three-echelon supply chain coordination with production strategy with advance selling target. Prod. Oper. Manage. 27 (7),
a loss-averse retailer and revenue sharing contracts. Int. J. Prod. Econ. 1221–1235.
Hu, P., Shum, S., Yu, M., 2015. Joint inventory and markdown management for Weraikat, D., Zanjani, M.K., Lehoux, N., 2016. Two-echelon pharmaceutical reverse
perishable goods with strategic consumer behavior. Oper. Res. 64 (1), 118–134. supply chain coordination with customers incentives. Int. J. Prod. Econ. 176, 41–52.
Johnson, G.D., Thompson, H.E., 1975. Optimality of myopic inventory policies for Wu, X., Zhang, F., 2014. Home or overseas? An analysis of sourcing strategies under
certain dependent demand processes. Manage. Sci. 21 (11), 1303–1307. competition. Manage. Sci. 60 (5), 1223–1240.
Lei, Q., Chen, J., Wei, X., Lu, S., 2015. Supply chain coordination under asymmetric Yan, H., Liu, K., Hsu, A., 2003. Order quantity in dual supply mode with updating
production cost information and inventory inaccuracy. Int. J. Prod. Econ. 170, forecasts. Prod. Oper. Manage. 12 (1), 30–45.
204–218. Yu, M., Ahn, H.-S., Kapuscinski, R., 2014. Rationing capacity in advance selling to
Li, X., Lian, Z., Choong, K.K., Liu, X., 2016. A quantity-flexibility contract with signal quality. Manage. Sci. 61 (3), 560–577.
coordination. Int. J. Prod. Econ. 179, 273–284. Yu, M., Debo, L., Kapuscinski, R., 2015a. Strategic waiting for consumer-generated
Lovejoy, W.S., 1990. Myopic policies for some inventory models with uncertain demand quality information: Dynamic pricing of new experience goods. Manage. Sci. 62
distributions. Manage. Sci. 36 (6), 724–738. (2), 410–435.
Ma, G., Liu, Q., Wu, L., Chen, E., 2015. Identifying hesitant and interested customers Yu, M., Kapuscinski, R., Ahn, H.-S., 2015b. Advance selling: Effects of interdependent
for targeted social marketing. In: Pacific-Asia Conference on Knowledge Discovery consumer valuations and seller’s capacity. Manage. Sci. 61 (9), 2100–2117.
and Data Mining. Springer, pp. 576–590.

11

You might also like