Integrated Fleet and Demand Control for On-Demand Meal Delivery Platforms

• 成果类型: Article; Early Access
• 署名作者: Hildebrandt, F. D.; Lesjak, Z.; Strauss, A.; Ulmera, M. W.
• 署名单位: Otto von Guericke University; WHU - Otto Beisheim School of Management
• 刊物名称: MANAGEMENT SCIENCE
• ISSN/ISSBN: 0025-1909
• DOI: 10.1287/mnsc.2022.02039
• 发表日期: 2025
• 关键词: instant delivery; stochastic dynamic vehicle routing; demand management; display optimization; Reinforcement Learning

摘要

We show how integrated fleet and demand control can be effectively used to benefit all stakeholders in on-demand restaurant meal delivery. Fleet control-that is, the assignment of orders to couriers-is the main control mechanism to steer delivery operations. Another, mostly overlooked, control mechanism is demand control via display optimization-that is, the ordering of restaurants' display positions on the meal delivery platform. Based on historical customer interactions with a meal delivery platform, we reveal that display positions have a major effect on customers' restaurant choices. We then leverage this effect by proposing an integrated, scalable reinforcement learning approach that simultaneously optimizes fleet and demand control. We employ our solution method on simulations of large-scale on-demand meal delivery operations with endogenous customer behavior to derive managerial insights on the value of integrated fleet and demand control. Our results demonstrate that integrated fleet and demand control reduces delays experienced by customers, allows for more services per driver, decreases total travel time per driver, guarantees fresher meals, and provides equal opportunities for all participating restaurants. Our results further highlight that selling display positions may cause operational inflexibility and, therefore, may cause significant delays in the fulfillment process. Finally, we show that careful display optimization not only improves service quality, but also platform revenue.