Supply Chain Contracts with Multiple Retailers in a Fuzzy Demand Environment

This study investigates supply chain contracts with a supplier and multiple competing retailers in a fuzzy demand environment. The market demand is considered as a positive triangular fuzzy number. The models of centralized decision, return contract, and revenue-sharing contract are built by the method of fuzzy cut sets theory, and their optimal policies are also proposed. Finally, an example is given to illustrate and validate the models and conclusions. It is shown that the optimal total order quantity of the retailers fluctuates at the center of the fuzzy demand. With the rise of the number of retailers, the optimal order quantity and the fuzzy expected profit for each retailer will decrease, and the fuzzy expected profit for supplier will increase.


Introduction
Over the last decade or so, supply chain management has emerged as a key area of research among the practitioners of operations research.In recent years, coordination mechanism of supply chain contracts has become one of the most challenging issues facing both practitioners and scholars.Supply chain contracts such as return contract and revenuesharing contract are instruments for supply chain coordination, which shift the uncertain demand from the retailer to the supplier, thus encouraging the retailer to increase order quantities.
A large body of literature has explored to coordinate the supply chain with return contract and revenue-sharing contract during the last two decades.Pasternack [1] first claimed that an appropriate return policy can fully coordinate a single-supplier single-retailer supply chain, which was then extended by Mantrala and Raman [2] to the situation where the retailer had several stores.Taylor [3] and Lee et al. [4] studied the return contract with effort-dependant demand.They showed that in this problem, it attained supply chain coordination combined with feedback policy.Yao et al. [5] analyzed the profits of both actors when the manufacturer and retailer shared or did not share the forecast information in returns policy.Yue and Raghunathan [6] discussed the impact of a full return policy as well as information sharing on the manufacturer and the retailer under information asymmetry.Bose and Anand [7] considered the wholesale price as an exogenous price to study returns policies for coordinating the supply chain.They showed that, in general, an equilibrium returns policy was not Pareto efficient with respect to a price-only contract, but when the wholesale price was sufficiently high, the equilibrium returns policy was Pareto efficient.These conclusions were consistent with those of Yao et al. [8].Ding and Chen [9] studied the return contract issues of a three-level supply in a single-period model.Yao et al. [10] analyzed the impact of price-sensitivity factors on characteristics of return contract in a single-period product supply chain.Mollenkopf et al. [11] used an empirical study to explore how internet product returns management systems affect loyalty intentions.Chen and Bell [12,13] showed that the customer returns affect the firm's pricing and inventory decision and proposed an agreement between the manufacturer and the retailer that includes two buyback prices.Chen [14] proposed a returns policy with a wholesale-pricediscount scheme that can achieve supply chain coordination.Ai et al. [15] analyzed the implementation of full returns policies in the chain-to-chain competition.

Mathematical Problems in Engineering
Revenue-sharing contract has been applied in the video cassette rental and movie industry with much success.Giannoccaro and Pontrandolfo [16] showed that revenue-sharing could coordinate members in the newsboy channel with three stages: supplier, manufacturer, and retailer.Cachon and Lariviere [17] intensively discussed a revenue-sharing contract between a single supplier and a single retailer in a singleperiod newsboy problem.Gupta and Weerawat [18] designed a revenue-sharing contract to maximize the centralized revenue by choosing an appropriate inventory level.Yao et al. [19] investigated a revenue-sharing contract for coordinating a supply chain comprising one manufacturer and two competing retailers.Linh and Hong [20] studied a revenue-sharing contract in a two-period newsboy problem.Van Der Rhee et al. [21] proposed a revenue-sharing mechanism in multiechelon supply chains.Ouardighi and Kim [22] considered a single supplier collaborating with two manufacturers on designing quality improvements for their respective products under a revenue-sharing contract.Krishnan and Winter [23] studied the role of revenue-sharing contracts in supply chains and established a foundation in aligning incentives.Sheu [24] explored revenue-sharing contracts under price promotion to end customers with three types of promotional demand patterns.Zhang et al. [25] investigated a revenuesharing contract with demand disruptions in a supply chain comprising one manufacturer and two competing retailers.Palsule-Desai [26] studied revenue-dependent contracts and revenue-independent contracts in a two-period model, and they showed that both types of revenue-sharing contracts could coordinate the supply chain; however, there existed situations in which revenue-dependent contracts outperformed revenue-independent contracts.The conventional studies have focused on the cases in which the demands are probabilistic.In other words, the demands follow certain distribution function.However, in practice, especially for new products, the probabilities are not known due to lack of history data.In this case the demands are suitably described subjectively by linguistic terms, such as "high, " "low, " or "approximately equal , but definitely not less than  and not greater than ." Thus, the uncertain theory, rather than the traditional probability theory, is well suited to the supply chain models problem.Therefore, we assume that the external demand can be approximately forecasted and expressed as a triangular membership function.
In this paper, the demands are approximately estimated by experts and regarded as fuzzy numbers.Return contract and revenue-sharing contract with multiple competing retailers in a fuzzy demand environment will be discussed, and the impact of the supplier's production cost and the number of retailers on the models will be analyzed.
The rest of the paper is organized as follows.Section 2 introduces some definitions and propositions about fuzzy set theory and notations related to this paper.Section 3 develops three fuzzy supply models with multiple competing retailers.Section 4 provides a numerical example to illustrate the result of the proposed contracts.Section 5 summarizes the work done in this paper.

Fuzzy Set Theory
Definition 1.The fuzzy set Ã = ( 1 ,  2 ,  3 ), where  1 <  2 <  3 and defined on , is called the triangular fuzzy number, if the membership function of Ã is given by where  1 and  3 are the lower limit and upper limit, respectively, of the triangular fuzzy number Ã.For  ∈ [ where Ã−1  () and Ã−1  () are, respectively, the left and right boundaries of Ã(), with Ã−1  () = inf { ∈  :  Ã() ≥ } , Ã−1  () = sup { ∈  :  Ã() ≥ } . ( Example 4. For any  ∈ [0, 1], the  cut set of a triangular fuzzy number Ã = ( 1 ,  2 ,  3 ) is Based on the extension principle in fuzzy sets, we have the following Propositions 5 and 6.Proposition 5.For any  ∈ [0, 1], let Ã be a positive triangular fuzzy number and let  be a nonzero real number; then respectively, be the  cut set of the positive triangular fuzzy numbers B and C; then Proposition 7 (see B. Liu and Y.-K.Liu [27]).Let Ã be a positive triangular number; the expected value of Ã is Proposition 8 (see Y.-K.Liu and B. Liu [28]).Let Ã and B be two independent positive triangular fuzzy numbers with finite expected values.Then for any real numbers  and , one has 2.2.Problem Descriptions.Consider a single-period setting for a two-echelon supply, consisting of a supplier and multiple competing retailers with fuzzy demand.We assume that at the beginning of the selling season, the retailer  ( = 1, 2, . . ., ) has no inventory on hand and must decide the order quantity   from the supplier.Then, the retailer  sells his order of short-life products, such as personal computers, consumer electronics, or fashion items, with high uncertain demand.
The products are sold only in one period.As the lead times of such goods are much longer than their selling season, the actors have no chance to place a second order.We consider the total uncertain demand faced by the retailers as a positive triangular fuzzy variable D = ( 1 , 2 ,  3 ) with the most possible value  2 , where 0 <  1 <  2 <  3 .The fuzzy demand D means that the total demand is about  2 . 1 and  3 are the lower limit and upper limit, respectively, of the fuzzy demand D and described by a general membership function  D(): Let the total retail demand be divided between the  retailers proportional to their stocking quantity; that is, retailer 's demand, D , is where  = ∑  =1   .The following notations are used for a product in the models: : the retail price; : the wholesale price; : the per unit product cost incurred to the supplier; : the return price offered by supplier in return contract; Φ: the fraction revenue of the retailer in revenuesharing contract and 0 < Φ < 1; Π : the fuzzy profit of the supplier; Π  : the fuzzy profit of the retailer ; ΠSC : the fuzzy profit of the supply chain.
The supplier and the retailer  are assumed to be risk neutral and pursued maximization of their fuzzy expected profits.

Centralized Decision
Making with Fuzzy Demand.Consider a supply chain occupied by an integrated actor, which can also be regarded as the retailers and the supplier-making cooperation.The fuzzy profit of two-stage supply chain can be expressed as Since the fuzzy demand D = ( 1 ,  2 ,  3 ) in ( 11) is a positive triangular fuzzy number, we know that the order quantity  has two cases; that is, Theorem 9. When  <  ≤ 2, the optimal total order quantity  * of the retailers is Proof.
(a) For  ∈ [0, ()], the  cut set of the supply chain's fuzzy profit is (b) For  ∈ ((), 1], the result turns to By (7), we can get the fuzzy expected profit [ ΠSC ] as The first and second derivatives of [ ΠSC ] in ( 16) can be obtained as follows: Since () is an increasing function with   () > 0, therefore d 2 [ ΠSC ]/d 2 is negative and [ ΠSC ] is concave in .
Hence, the optimal total order quantity of the retailers can be obtained by solving d[ ΠSC ]/d = 0, which gives Since 0 < ( * ) ≤ 1, thus we can get  <  ≤ 2.
The proof of Theorem 9 is completed.
Theorem 10.When  > 2, the optimal total order quantity  * of the retailers is Proof.If  ∈ ( 2 ,  3 ], then the  cut set of min{, D} is (a) For  ∈ [0, ()], the  cut set of the supply chain's fuzzy profit is (b) For  ∈ ((), 1], the result turns to By (7), we can get the fuzzy expected profit [ ΠSC ] as The first and second derivatives of [ ΠSC ] in (23) can be obtained as follows: Since () is a decreasing function with   () < 0, therefore d 2 [ ΠSC ]/d 2 is negative and [ ΠSC ] is concave in .
Hence, the optimal total order quantity of the retailers can be obtained by solving d[ ΠSC ]/d = 0, which gives Since 0 < ( * ) ≤ 1, thus we can get  > 2.
The poof of Theorem 10 is completed.
From ( 16) and ( 23), Theorems 9 and 10, we can easily obtain the optimal fuzzy expected value of the profit for the integrated supply chain, which is given by 3.2.Return Contract with Fuzzy Demand.In a return contract, the supplier sets a wholesale price  and gives the retailer  ( = 1, 2, . . ., ) a return price  for unsold products at the end of the season.The fuzzy profit of the retailer  can be expressed as follows: where  = ∑  =1   and  − =  −   .The retailer  ( = 1, 2, . . ., ) tries to maximize its fuzzy expected profit [ Π  (  ,  − )] in return contract by choosing the optimal order quantity   , which solves the following model: Theorem 11.When  <  ≤ 2, the optimal wholesale price  * in return contract is (a) For  ∈ [0, ()], the  cut set of the retailer 's fuzzy profit is (b) For  ∈ ((), 1], the result turns to By (7), we can get the fuzzy expected profit [ Π  (  ,  − )] as The first and second derivatives of [ Π  (  ,  − )] in (33) can be obtained as follows: Since () is an increasing function with   () > 0 and  > , therefore d 2 [ Π  (  ,  − )]/d 2  is negative and [ Π  (  ,  − )] is concave in   .Hence, there exists an optimal order quantity for retailer  for each  − , where  − = ∑  =1, ̸ =    .A set of order quantity  * = ( * 1 , . . .,  *  ) is a Nash equilibrium of the decentralized system if each retailer's order quantity is a best response.Thus, any Nash equilibrium must satisfy each retailer's first-order condition.Let d[ Π  (  ,  − )]/d  = 0; we can get Equation ( 35) gives each retailer's equilibrium order conditional on  * being the equilibrium total order quantity.Hence, (35) describes an equilibrium only if  * =  *  .Substitute (35) into  * =  *  and simplify In order to fully coordinate the supply chain, let ( * ) = 2(− )/; we can obtain The poof of Theorem 11 is completed.
Theorem 13.In fuzzy return contract, the retailer  ( = 1, 2, . . ., ) and the supplier attain their optimal fuzzy expected value of the profits at  * in fuzzy return contract, where Proof.Consider the following.

Revenue-Sharing Contract with Fuzzy
Demand.In a revenue-sharing contract, the retailer  ( = 1, 2, . . ., ) shares with the supplier a percentage of his revenue.Let (1 − Φ) be the fraction the supplier earns, and then Φ is the fraction the retailer  keeps.
Thus, we can express the fuzzy profit of the retailer  as follows: where  = ∑  =1   and  − =  −   .The retailer  ( = 1, 2, . . ., ) tries to maximize its fuzzy expected profit [ Π  (  ,  − )] in revenue-sharing contract by choosing the optimal order quantity   , which solves the following model: Theorem 14.When  <  ≤ 2, the optimal wholesale price  * * in revenue-sharing contract is Proof.

Numerical Example
In this section, we tend to further elucidate the previously proposed two contracts with a numerical example.We will analyze the effect of the number of retailer  on the other parameters.Suppose the most possible value of the demand in market is  2 = 2000 units and the maximum and minimum possible values of the demand are, respectively,  1 = 1000 units and  3 = 3000 units; that is to say, the fuzzy demand is D = (1000, 2000, 3000).Let  = 100.00$per unit.
The optimal order quantity, wholesale price, and fuzzy expected profit of the actors in return contract and revenuesharing contract can be listed in Tables 1 and 2, respectively.
From Tables 1 and 2, we analyze the influence of parameters  and  on the optimal equilibrium values as shown below.
(a) It is obvious that the optimal order quantity for retailer  will decrease along with the rise of the supplier's production cost  and the number of retailers  when the other parameters are fixed.Particulary, in this numerical example, the optimal total order quantity  * is equal to the most possible value of fuzzy demand when  = 50.When  > 50 and  < 50, the optimal total order quantity of retailer is located at the left and right of the most possible value of fuzzy demand D, respectively.The optimal wholesale price will increase along with the supplier's production cost  and thenumber of retailers  when the other parameters are fixed.
(b) From Tables 1 and 2, it can be noted that when the other parameters are fixed in both contracts, the fuzzy expected profit of the retailer  will decrease along with the rise of parameters  and .On the other hand, the fuzzy expected profit of the supplier will decrease along with the rise of  and increase along with the rise of .

Conclusions
This paper formulates supply chain contracts based on fuzzy set theory, where a supplier and multiple competing retailers adopt return contract and revenue-sharing contract.In order to examine models performance in fuzzy demand, we use fuzzy cut sets method to solve this problem.The advantage of the proposed method is that it removes the need for enumeration over alternative values and provides a better understanding of the relations among the decision parameters.The technique proposed in this paper is easier to implement and requires less data.It is appropriate when the environment is complex, ambiguous, or there is a lack of statistical data.