Data-driven multi-location inventory placement in digital commerce

Digital commerce has become an indispensable part of global retail. Digital commerce retailers usually build large logistics networks with multiple distribution centers (DCs) to serve widespread consumers. In this paper, we study multi-location inventory placement for online retailers to fulfill customer demands. Specifically, we consider three decision-making problems: (i) in which DCs to place inventory, (ii) how to set base-stock levels for inventory-holding DCs, and (iii) from which DCs to fulfill customer demand. The main challenge is to achieve the optimal trade-off between inventory cost savings from inventory pooling and the increased demand fulfillment cost associated with placing inventory far from consumers. To investigate the trade-off, we propose a data-driven stochastic program under two different demand fulfillment policies, namely fixed and virtual pooling. We evaluate the effectiveness of the proposed method through a case study based on a real-world data set by a logistics company. The proposed method achieves an average cost reduction of 19.2% compared to the company’s current inventory placement policy. Further, we conduct ABC-XYZ analysis for more than 7,700 stock keeping units (SKUs) in the data set. The comparison of inventory placement decisions between different SKU categories suggests that digital commerce retailers should place more inventory in local DCs for SKUs with steadily high demand rates and pool more inventory at central DCs for SKUs with low demand rates and high variance. Additionally, we perform a systematic sensitivity analysis with controllable problem parameter configurations to investigate the impact of different parameters on inventory placement decisions.