Optimizing Task Generation and Assignment in Crowdpicking

Abstract

As omnichannel operations become increasingly important for meeting diverse customer expectations in retail, continuous innovation in service and business models is essential to maintain a competitive edge. While effective order fulfillment is key to omnichannel success, the manual picking process in physical stores, one of the major driver of fulfillment costs, still offers substantial opportunities for improvement. This article focuses on the crowdpicking model as an innovative approach to manage online order picking operations in physical stores by leveraging existing in-store customers, offering a business model with considerable potential. We explore the real-time assignment of orders to in-store customers using machine learning to identify effective assignment policies. These policies are combined with a task-decomposition strategy to reduce picking costs and enhance crowdpicker participation as a key resource. The proposed crowdpicking model and its real-time management framework are tested on real-world data. Our results show that a well-managed crowdpicking system can lower order picking costs by more than 20% and provide actionable insights for managers in designing such systems.