Information control and usage for Pareto improvement of supply chains

Hung Tuan Do, Purdue University

Abstract

In Chapter 1, we study the problem of using manufacturer queue information with order placement control to provide Pareto improvement in a decentralized supply chain. Motivated by examples in industry with heterogeneous information availability in decentralized supply chains, we model a supply chain with one manufacturer (or assembler) and two retailers (or OEMs) who order independently. The manufacturer may provide her queue state information to one (informed) retailer with possible constraints on order placement. We propose an order placement constraint imposed by the manufacturer and a conforming order policy class for the informed retailer that may potentially increase the retailer's order flow variability, but, by being negatively correlated with queue length, stochastically decreases the manufacturer queue's length and variability. Within this order policy class, we propose a specific order policy that is guaranteed to make the informed retailer not worse off while making the other retailer better off and the manufacturing queue stochastically smaller and less variable, hence beneficial to the manufacturer, and thus providing Pareto improving outcomes. We provide a general analysis and a computational method to obtain stationary distributions of the queue size and thereby costs of the retailers for any given pure, Markov and stationary order policy. We propose a non-preemptive priority scheduling to permit different splits of the benefit across the two retailers. Numerical and simulation results confirm and demonstrate the magnitude of these cost reductions. When both retailers have access to the information associated with the proposed constraint and implement the proposed ordering policy, our simulations show that inducing a negative correlation between the order flow and the manufacturer state decreases the cost of each entity even further. This model is thus a building block towards understanding how information sharing can be coupled with order management to improve outcomes for participants in a supply chain. In Chapter 2, we study the problem of using manufacturer queue information with pricing control policies to provide Pareto improvement in a decentralized supply chain. Our research problems are motivated by the scenarios in industry with information availability in decentralized supply chains, we model a supply chain with one manufacturer (or assembler) and two retailers (or OEMs) who order independently using a base stock order policy. The manufacturer queue state information can be provided to one or both retailers. The case of no information sharing is used as a benchmark in this paper. We propose and analyze the promotion policies via pricing, given the information availability, that provide cost saving to each business entity and thereby Pareto improvement to the supply chain. We establish analytically that the number of outstanding orders of each retailer is stochastically reduced and less variable when both retailers have the queue state information and employ our proposed promotion policies. This results in a lower cost for each retailer. When only one retailer is informed, we show that the total queue is stochastically smaller and less variable, resulting in a lower cost for the uninformed retailer. Using approximate analysis, we show that the cost of the informed retailer is also decreased. Finally, our numerical calculation and simulation results demonstrate that the cost of each retailer is indeed significantly reduced, monotonically in promotion level. Our results challenge a popular belief for a similar scenario that promotion of retailers should not be synchronized. The problems in Chapter 3 are motivated by the procurement problems faced by a manufacturer of finished products that uses commodity metals as raw material. We study the optimal procurement and inventory control policies when the raw material can be procured from two sources, i.e., fixed period swing contracts and spot markets, in order to satisfy a stochastic demand. Under a swing contract, the company makes a commitment on the total purchase quantity over a fixed contract interval. The contract price is specified and fixed throughout the term, but the quantity delivered in each period of the contract is flexible. Given that the spot price evolves stochastically, the company may want to use a swing contract to hedge against the uncertainty of the spot market price and at the same time take advantage when low spot prices are realized. We model the problem as a finite horizon, multi-period stochastic dynamic program in which the company makes decisions regarding the swing contract quantity and then, at the beginning of each period, dynamically chooses the quantities purchased from swing contract and spot market. We use data from the manufacturer to numerically compare the impact of the optimal policies against that of the policies implemented currently. Our numerical results show that by applying the proposed policy, the company can obtain, on average, more than 8.89% cost saving, compared with their current policies. (Abstract shortened by UMI.)

Degree

Ph.D.

Advisors

Iyer, Purdue University.

Subject Area

Management|Operations research

COinS