Safety Stock Allocation in an Online Retailing Network: A Stochastic Optimization Approach

Abstract

Online retailers utilize lateral transshipment to meet the immediate shortage of demand at one location from surplus at another location within their retailing network. One option to reduce such potentially expensive transshipment is to hold additional safety stock at each location. We investigate the allocation of safety stock in a network of fulfillment centers with the possibility of lateral transshipment. Our problem setting is motivated by online retailers who operate multi-node retailing networks of fulfillment centers under a periodic review and order-up-to policy. We formulate the safety stock allocation problem as a stochastic optimization model with uncertain demand, and optimally solve the model for a network with up to six fulfillment centers. The optimization model quickly becomes intractable, with an increase in problem size. Therefore, for larger networks, we propose a novel hub-and-spoke approach where the fulfillment centers are first clustered and then connected as a Minimal Spanning Tree, followed by a system-wide safety stock allocation via stochastic optimization. The hub-and-spoke approach performs quite favorably in our extensive computational experiments, and demonstrates the potential to save millions of dollars in outbound shipping costs for an online retailer. Our study provides online retailers with an easily implementable and efficient safety stock allocation mechanism across a complex and geographically distributed network of fulfillment centers.

Keywords: supply chain, safety stock, inventory, lateral transshipments, stochastic demand