Improving front distribution center fulfillment rates: a distributionally robust approach

In E-commerce distribution networks, front distribution centers (FDCs) are extensively employed to reduce delivery time, which has a significant impact on customers’ purchase intentions and loyalty. Upon customer order placement, the corresponding FDC promptly fulfills the order, ensuring a short delivery time. If there is shortage in the FDC, the order is then fulfilled by the regional distribution center (RDC) with a longer delivery time. Otherwise, a lost sale occurs. A key performance metric is the FDC fulfillment rate which reflects the proportion of orders successfully fulfilled by FDCs. In this paper, we design a distributionally robust allocation model that improves FDCs’ fulfillment rates and, at the same time, maintains the region’s overall fulfillment rate. We transform this model into an equivalent mixed integer second order conic programming (MISOCP) model, and an approximate mixed integer linear programming (MILP) model by partitioning the robust domain. Through numerical experiments, we investigate the impact of the balance coefficient on fulfillment rates and demonstrate the excellent performance of our model in a rolling horizon setting, particularly when faced with inaccurate demand forecasts. We implement our model within the distribution network of the home appliance industry of Tmall platform (the largest E-commerce retail platform in China), resulting in a notable improvement in the FDC fulfillment rate (exceeding 10%) and a substantial boost in gross merchandise volume (GMV).