Ximin Huang

Inducing high service capacities in outsourcing via penalty and competition

In this article, we consider a model of multiple make-to-order suppliers, which compete based on their promised sojourn times. A fixed penalty is imposed on the suppliers for failing the promised delivery time requirement at an agreed service level. The demand is allocated to the suppliers according to the promised sojourn times, using either supplier-selection (SS) approach or supplier-allocation (SA) approach as proposed in Benjaafar et al. (2007, Outsourcing via service competition, Management Science, 53 (2), 241–259). Their framework and results are then applied to obtain the competition outcome. It was found that the buyer can orchestrate the competition to induce the suppliers to offer lower promised sojourn times and select higher service capacities through restricting the suppliers to choose promised sojourn times above a certain level. The SS approach is also found to induce a better service quality than the SA approach. The role of the penalty level is then addressed. It was found that a higher penalty level induces a higher service capacity and is always beneficial to the buyer.

Higher-Order Markov Chains

Data sequences or time series occur frequently in many real world applications. One of the most important steps in analyzing a data sequence (or time series) is the selection of an appropriate mathematical model for the data. This is because it helps in predictions, hypothesis testing and rule discovery.

On improving incentive in a supply chain: Wholesale price contract vs quantity dependent contract

In this paper, we first study the performance of a supply chain consisting of one retailer and one supplier. The supplier sets the price scheme of some goods and the retailer then decides the order level and sells the goods in the market. Specifically, a quadratic cost function is assumed here to approximate the U-shape cost curve commonly observed in industries. Two kinds of contracts offered by the supplier are investigated, namely wholesale price contract and quantity dependent contract. Wholesale price is fixed under the first contract but varies depending on order level under the second one. We show that certain wholesale price contract successfully induces the retailer to order at a level such that supply chain profit is maximized, but extra cost in implementation may occur due to suppliers disagreement on this price. Given this, we propose an efficiency measure to show to what extent the wholesale price contract helps to increase supply chain profit. For quantity dependent contract, we show that it can coordinate the supply chain and leads to a proportional division of supply chain profit. We then generalize the analysis to cover the case of multiple retailers and single supplier where similar results are also obtained.

Quantity discount contract for supply chain coordination with false failure returns

Consumer return attracts more and more academic attention due to its rapidly expanding size, and a large portion of it falls into the category of false failure return, which refers to return without functional defect. In this paper, we exclusively consider profit results from exerting costly effort to reduce false failure returns in a reverse supply chain consisting of a retailer and a supplier. The supply chain as a whole has strong incentive to reduce false failure returns because it can avoid much reprocessing cost associated. But typically, retailers enjoy a full credit provided by suppliers in case of returns, and hence they may not have sufficient incentives to exert enough effort for supply chain profit maximization. In some scenarios they may even have the motivation to actually encourage such returns. We suggest using a coordination contract to resolve such profit conflicts. The contract we propose is a quantity discount contract specifying a payment to the retailer with an amount exponentially decreasing in the number of false failure returns. We give explicit forms of such contracts given different assumptions about distribution of the number of returns and we also prove that such contract is capable of increasing both retailers and suppliers profit simultaneously. Besides, when the contract is used together with other forward supply chain coordination contracts in a closed-loop chain, it is shown that it can act to deter retailers potential incentive to encourage false failure returns. Moreover, some modifications of the contract may lead to easy allocation of incremental profit within the supply chain.

Inducing optimal service capacities via performance-based allocation of demand in a queueing system with multiple servers

In this paper, we study the use of performance-based allocation of demand in a multiple-server queueing system. The same problem with two servers have been studied in the literature. Specifically, it has been proposed and proved that the linear allocation and mixed threshold allocation policies are, respectively, the optimal state-independent and state-dependent allocation policy in the two-server case. The multiple-server linear allocation has also been shown to be the optimal state-independent policy with multiple servers. In our study, we focus on the use of a multiple-server mixed threshold allocation policy to replicate the demand allocation of a given state-independent policy to achieve a symmetric equilibrium with lower expected sojourn time. Our results indicate that, for any given multiple-server state-independent policy that prohibits server overloading, there exists a multiple-server mixed threshold policy that gives the same demand allocation and thus have the same Nash equilibrium (if any). Moreover, such a policy can be designed so that the expected sojourn time at a symmetric equilibrium is minimized. Therefore, our results concur with previous two-server results and affirm that a trade-off between incentives and efficiency need not exist in the case of multiple servers.

Queueing Systems and the Web

In this chapter, we first discuss some more Markovian queueing systems. The queueing system is a classical application of continuous Markov chains. We then present an important numerical algorithm based on the computation of Markov chains for ranking webpages. This is a modern application of Markov chains though the numerical methods used are classical.

Manufacturing and Re-manufacturing Systems

In this chapter, we consider the application of the Markovian queueing systems discussed in Chap. 2 in modeling manufacturing systems and re-manufacturing systems. We adopt Hedging Point Production (HPP) policy as a production control policy. We note that in a queueing system, there are servers, customers, and waiting spaces. To model a make-to-order manufacturing system by a queueing system, one may regard a server as a machine. The customers can be regarded as the inventory of product or the jobs to be processed respectively; see for instance Buzacott and Shanthikumar [34]. In a manufacturing system, a certain amount of inventory (called the hedging point) is kept to cope with the fluctuation of demand and therefore production control is necessary. The system will stop production when this level of inventory is attained.

A Hidden Markov Model for Customer Classification

In this chapter, a new simple Hidden Markov Model (HMM) is proposed. The process of the proposed HMM can be explained by the following example.

Markov Decision Processes for Customer Lifetime Value

In this chapter a stochastic dynamic programming model with a Markov chain is proposed to capture customer behavior. The advantage of using Markov chains is that the model can take into account the customers switching between the company and its competitors. Therefore customer relationships can be described in a probabilistic way, see for instance Pfeifer and Carraway [170]. Stochastic dynamic programming is then applied to solve the optimal allocation of the promotion budget for maximizing the Customer Lifetime Value (CLV). The proposed model is then applied to practical data in a computer services company.

Hidden Markov Chains

Hidden Markov models (HMMs) have been applied to many real-world applications. Usually HMMs only deal with the first-order transition probability distribution among the hidden states, see for instance Sect.1.4. Moreover, the observable states are affected by the hidden states but not vice versa. In this chapter, we study both higher-order hidden Markov models and interactive HMMs in which the hidden states are directly affected by the observed states. We will also develop estimation methods for the model parameters in both cases.