Please use this identifier to cite or link to this item: http://hdl.handle.net/10397/91765
Title: Three essays in operations management : inventory management with demand learning, quality signaling and Bayesian persuasion in service
Authors: Luo, Zhenwei
Degree: Ph.D.
Issue Date: 2021
Abstract: The information uncertainty is quite common for decision makers in operations management. This thesis includes three different settings with information uncertainty, where the Bayesian updating framework is adopted as a parametric learning approach. In the first topic, we consider that an airline company offers early-bird-discount seats to customers and aims to maximize the expected profit via optimally allocating the seats for discount sales. The beliefs on demand parameters and buy-up substitution probability are updated using demand observations considering an "exploration-exploitation" tradeoff. Classic literature finds that unobservability of lost sales is a driving force for the "stock more" result, namely the Bayesian-optimal inventory level shall be kept higher than the myopic one to allow a better observability of demand. In contrast, we find that one can infer some information about lost sales from the substitution behavior of unsatisfied customers and hence may "stock less". We also find that to better observe the primary demand for the regular-price seat, one shall "stock more" discounted seats to reduce the chance of tangling the substitution demand with the primary demand. And, to better observe the substitution probability, one shall "stock less" discounted seats to observe substitutions. In the second topic, we consider a single-server queueing system whose service quality is either high or low. The server knows the actual quality level, and can signal it to customers via revealing or concealing his queue length. A signaling game is formed, and we adopt the sequential equilibrium concept to solve our game and apply the perfect sequential equilibrium as an equilibrium-refinement criterion. Under a general scenario in which the market is composed of both quality informed and uninformed customers, the unique equilibrium outcome is a pooling strategy when the market size is either below a lower threshold or above an upper threshold. And the separating equilibria may exist only when the market size falls between these two thresholds, under which uninformed customers can fully infer the server's quality type based on his queue disclosure behavior. In the third topic, we study a server's best queue-disclosure strategy in a single-server service system with uncertain quality level. We consider this problem as a Bayesian persuasion game. The server can commit to a strategy that states whether or not the queue length will be revealed to customers upon their arrival, given a realized quality level. We reformulate the server's decision problem as looking for the best Bayes-plausible distribution of customers' posteriors on service quality, which can be solved via a geometric approach. We also show that when the market size is sufficiently small (resp. large), the server always conceals (resp. reveals) the queue regardless of the realized service quality. In a medium-sized market, however, we numerically find that the server's optimal commitment strategy is often hybrid or mixed, that is, randomized over queue disclosure and concealment. We also extend our analysis to another scenario where the server is a social planner.
Subjects: Production management
Uncertainty (Information theory)
Hong Kong Polytechnic University -- Dissertations
Pages: xiv, 146 pages : color illustrations
Appears in Collections:Thesis

Show full item record

Page views

12
Citations as of May 22, 2022

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.