Service Network Design: Competing on Convenient Locations and Responsive Service Processes

We present a market-oriented service network design model with an attraction model of customer choice as the underlying demand model. We assume demand as a function of waiting time and distance, among other attributes. The service provider’s objective is to maximize profits, and the number of facilities, their locations, the design of the service process at each facility (in particular with regard to process capacity and service level) are the main decision variables. In the literature on stochastic location models with congestion, service facilities have typically been modeled as single stage queuing systems with single or multiple servers. We complement this work and model each facility as a multi-stage queuing network with multiple servers. This assumption fits well for many service facilities, e.g. food outlets (e.g. Vapiano, Starbucks etc.) and public facilities (e.g. bank, government offices etc.). We capture the uncertainties related to demand and service process at a facility through an ex ante simulation and derive performance measures related to the underlying queuing network with respect to possible demand and capacity configurations. The output is incorporated into the service network design model. Mathematically, the optimization model is nonlinear mixed-integer, and can be linearized using "big M" constraints and several auxiliary variables. Since practical applications may require to solve large-scale instances, we investigate heuristic solution procedures and present a case study of designing a new pizza delivery service in Germany