ASSORTMENT SELECTION FOR A FRONTEND WAREHOUSE : A ROBUST DATA-DRIVEN APPROACH

A frontend warehouse is aimed at offering same-day or even faster delivery to consumers for online retailers. Located in urban areas or even residential areas, frontend warehouses are typically very small due to the expensive rent and can accommodate only a very limited number of products. In order to improve the fulfillment rate directly from the frontend warehouse, we need to properly select the products for the warehouse in the face of future demand uncertainty. To address this challenge, firstly, given historical order data, we propose a prediction model combining exponential smoothing and community detection to forecast demand and structure of future orders. Second, to enhance robustness of the solution, we propose a robust optimization model and develop an efficient solution approach. Finally, we test our method on real-world data from Tmall and demonstrate its superiority over a few benchmarking methods.