Integrating customer choice in differentiated slotting for last-mile logistics

In this paper, we study an e-grocer's problem of differentiated slotting in last-mile logistics for attended home delivery. The purpose of differentiated slotting is to statically determine which time slots should be on offer throughout the selling horizon in each area of a delivery region on each future service day such that costeffective last-mile delivery schedules on an operational level can be expected. For this purpose, in addition to an appropriate approximation of potential delivery tours and costs, it is crucial to adequately consider the customer choice behavior between slots. In this paper, we propose a model-based, profit-oriented slotting approach that accounts for such choice behavior in a realistic, sophisticated fashion using a finite-mixture multinomial logit choice model. We formulate the resulting optimization problem as a non-linear mixedinteger program and show how it can be linearized. Further, we conduct a computational study to examine the benefit of the new approach. In particular, we demonstrate the superiority of our approach to existing approaches in the academic literature, which neglect or simplistically approximate customer choice behavior and which are purely cost-based.