Studies On Inventory Systems With Multiple Suppliers And/or Multiple Demand Classes

As a node in supply chain, an inventory system combines supply and demand. Inpractice, a firm procures from upstream suppliers as a customer, and provides down-stream customers products and service as a supplier. Based on the above two rolesthat a firm plays in a supply chain, this thesis studies inventory systems with multiplesuppliers and/or multiple demand classes.For inventory system with multiple suppliers, we consider a multi-period inven-tory model with replenishment orders required to be splitted according to a simplethumb rule. An exact algorithm with polynomial time and a heuristic are developed,which solve the inner-level optimal order allocation problem and the outer-level opti-mal ordering problem, respectively. Numerical experiments show that the heuristic isaccurate and e?cient. In all the experiments, relative errors do not exceed 1%.For inventory system with multiple demand classes, we first di?erentiate cus-tomers according to whether they accept backorders in case of shortage. For a modelwith two demand classes and no setup cost for ordering, we show that the classicalbase-stock policy is only partially optimal. For a model with multiple demand classesand setup cost, the optimal ordering policy is a state-dependent (s, S ) policy, whereasthe optimal inventory allocation policy is a state-dependent prioritization policy.Then, we di?erentiate customers according to whether their demand rate dependson inventory. Evidences from the retail industry show that increasing the displayedamount of inventory often leads to increase of demand rate. However, for customersfrom a non-retail channel, demand rate is unlikely to be a?ected by inventory. Wesolve the optimal inventory allocation problem between a retail channel and a non-retailchannel, and analyze how the optimal decision reacts when the dependency betweendemand and inventory varies.We also study the ordering behavior of a retailer with time-inconsistent prefer-ences. Under the assumption that random demands follow log-concave distributions, we show that the retailer's ordering decision follows a base-stock policy. Based onthis result, we further study a supplier's inventory management problem, with demandgenerated from several time-inconsistent retailers.After studying inventory systems with multiple suppliers and multiple demandclasses separately, we further consider a periodic review inventory system that pro-cures from two suppliers and allocates inventory to multiple demand classes. The twosuppliers di?er in order lead-time, while the customer classes are di?erentiated by theirimportance to the system. Theoretical analysis shows that the optimal order allocationquantity for the short lead-time supplier can be determined by a modified base-stockpolicy, and that for the long lead-time supplier is characterized by a state-dependentmodified base-stock policy. Moreover, the inventory allocation policy is a critical levelpolicy.