Dynamic inventory management with expediting

Abstract

In modern global supply chains, goods travel stochastically from suppliers to their final destinations through several intermediate installations such as ports and distribution facilities. In such an environment, the supply chain must be agile to respond quickly to demand spikes. One way to achieve this objective is by expediting outstanding orders from the intermediate installations through premium delivery. In this research, we study the optimal expediting and regular ordering policies of a serial supply chain with a radio frequency identification deployment at each installation. Radio frequency identification technology allows capturing the state of the system, i.e., the time and location of goods, at any point in time, and thus enables to expedite outstanding orders directly to the destination, which faces stochastic demand. We identify systems, called sequential, that yield simple and tractable optimal policies. For sequential systems, outstanding orders including expediting do not cross in time. For such systems, we find that the optimal policies of expediting and regular ordering are the base stock type policies. The directional sensitivity of the base stock levels with respect to expediting costs is also obtained. We provide an important managerial insight on the radio frequency identification technology: we need to actively use the additional information from the radio frequency identification technology through new business processes such as expediting to unveil more benefits from the supply chain. On the other hand, orders may cross in time for systems that are not sequential, thus in such a case optimal policies are hard to obtain. We propose a heuristic for such systems and discuss its performance and limitation.

(cont.) Lastly, as an extension to the model, we study the optimal policies of expediting and regular ordering when there is an expiry date on outstanding orders. The optimal expediting policy identifies a number of base stock levels depending on the age of the orders, but the structure of the optimal policy remains simple for sequential systems.