Abstract
Micro-fulfillment centers (MFCs) play a critical role in ensuring that retailers can effectively provide omnichannel shopping services (whether through ultra-fast home delivery or pickup), and an increasing number of retailers are establishing MFCs to avail of these benefits. Establishing an MFC is only part of the solution, as the advantages will materialize only if the MFCs are managed effectively. One of the challenges involved in doing so concerns the decisions around what products to keep in the MFCs, and in what quantities. In this paper, we propose an effective threshold-based policy to help retailers make well-informed decisions, by formulating the MFC inventory decision problem as a Markov decision process. Results from computational experiments conducted on both simulated data and real data from a retailer show that this policy, which accounts for stochastic demand and makes forward-looking decisions, is scalable and identifies near-optimal solutions very quickly compared to state-of-the-art benchmarks. In particular, the experiments on real data show that the problem can be solved in a few seconds even when hundreds of thousands of orders and thousands of items are involved, illustrating the practical viability of this approach.