Optimal information acquisition, inventory control, and forecast sharing in operations management

Efficient information acquisition and sharing is the key to unlocking latent values in many operations. In this thesis, I first examine an inventory management problem with unknown demand distribution and unknown substitution probability. The optimal inventory level depends upon both the demand distribution and substitution probability. Furthermore, the choice of inventory level influences the customer behavior we observe, and hence our learning about the demand distribution and substitution probability. I show that whether one should "stock more" or "stock less" so as to optimally acquire demand information depends critically on the perishability of inventory and observability of lost sales.; I then extend the study to an airline revenue management problem in which the manager dynamically controls inventory (seat capacity) to learn about customers' substitution behavior among multiple fare products. Estimating the customers' likelihood to substitute as well as their original purchase probability is crucial in determining the expected revenue-maximizing booking control policy. I show that the optimal Bayesian booking control policy for a two-fare model is a state-dependent base-stock policy. Simulation results further suggest that Bayesian booking control policy with observed substitution outperforms policies that do not account for active learning.; Finally, I study the forecast sharing problem in a two-level supply chain. Under a fairly general forecast evolution model, I quantify the magnitude of bullwhip effect reduction that results from forecast sharing. An effective way of sharing advance forecast information is to have the retailer place advance orders with the supplier. Placing advance orders generates advance demand information for the supplier and also eliminates the need for the supplier to guess the retailer's underlying ordering policy. I thus derive the optimal advance order quantity for the retailer, the quantity that optimally balances the need to respond to external demand changes and to reduce the bullwhip effect. In addition, I show that designing a proper transfer cost scheme between the supplier and the retailer improves the overall system performance.