Choice-based dynamic time slot management in attended home delivery

Abstract

E-grocers with an attended home delivery service model operate in a highly competitive market characterized by thin profit margins. To ensure a profit-maximizing delivery schedule, the requirements for the joint management of demand and the vehicle routes are substantial. Therefore, we study an e-grocer’s operational problem of managing demand by means of dynamic time slot allocation. The purpose of dynamically allocating time slots is to influence customers’ choices by offering a selection of time slots to a customer request, such that the overall expected profit of the resulting delivery schedule is maximized. The time slot offer decisions mainly depend on a request’s opportunity cost. Hence, we first propose a mixed-integer linear program to approximate this opportunity cost, which scales linearly with the number of decision variables. In this approximation, we consider the consequences of expected time slot offer decisions for future customers on the final delivery schedule. We explicitly incorporate customer choice behavior using a generalized attraction model. Second, we propose a non-linear binary program and its linearization based on the underlying choice model to make time slot offer decisions using the approximated opportunity cost. Due to the formulation’s structural properties, it can be efficiently solved. In a computational study, we show the superiority of our approach in comparison to benchmarks applied in the academic literature and its applicability in an online environment.