Chen, Xin (chair)
Wang, Qiong (director of research)
Seshadri, Sridhar (member)
Sun, Ruoyu (member)
Assortment Planning in Supply Chains
Three problems have been studied in this thesis. First, due to state regulations, beer and soft-drink manufacturers must sell their products through a single wholesaler in a geographical area in the U.S., which is referred to as a two-tier supply chain. Manufacturers compete by assortments and/or wholesale prices. The wholesaler sets the optimal market prices of products from all manufacturers, and the end consumers' choice behavior is governed by the Multinomial Logit (MNL) model. From the manufacturers' perspective, we formulate their assortment competition and joint pricing and assortment competition as Nash games, prove the structures of the best responses, and show the existence of a pure-strategy Nash equilibrium for each game.
Second, continuing with the two-tier supply chain, when a manufacturer sells multiple categories of products, or each manufacturer's band name effect cannot be ignored, we use the Nested Logit (NL) model to describe the consumers' purchase behavior. We study a monopoly setting where there is only one manufacturer, and provide efficient approaches to solve the assortment planning (and pricing) problem without or with cardinality constraints. We also discuss an oligopoly setting where multiple manufacturers compete by assortments (and wholesale prices), and determine the best responses and the existence of a pure-strategy Nash equilibrium in such games.
Third, assortment and inventory decisions are important for a retailer when frequently changing assortments is not possible. We consider a joint assortment and inventory planning problem when the deterministic part of the demand is modeled by the MNL model and the stochastic part is multiplicative. We prove the structure of the optimal assortment and the optimal order quantities when the information is perfect. When the distribution of the stochastic part is unknown, we determine the assortment and inventory solutions to maximize the worst-case profit. When the attraction factors in the deterministic part are from an uncertainty set, we provide the upper bound and the lower bound for the max-min problem.