Effective management policies for remnant inventory supply chains

Abstract

A remnant inventory distribution system is considered where a set of geographically dispersed distribution centers meet stochastic demand for a one-dimensional product. This demand arises from some other set of geographically dispersed locations. The product is replenished at the centers in a limited number of standard sizes, while the demand is for various smaller sizes of the product and arrives over time according to a Poisson process. There are costs associated with cutting and transportation and scrap can be profitably reclaimed. The combined production (cutting) and distribution problem is modeled as a network. A linear programming formulation is solved for a deterministic version of this problem using mean demand rates and the optimal dual multipliers are used to assign inherent values to remnants of various sizes. These values are then used to develop a price-directed policy that can be used in a stochastic environment. A simulation study shows that this policy significantly outperforms heuristic policies from the literature as well as other heuristic policies that have been used in the steel industry for a similar problem. Theoretical insights into the structure of the proposed optimization problem are provided along with proofs of several important results.