A multi-stage approach for optimizing the three-echelon joint newspaper distribution network of two major publishers in São Paulo, Brazil

Abstract

In this paper we address an optimization problem that arises in two- or three-echelon distribution network design whereby there is an interplay between the production and transfer schedule of semi-customizable products and their delivery under tight time windows. It is motivated by a real-world problem of optimizing the joint newspaper last-mile distribution network for SPDL (São Paulo Distribution and Logistics), a dedicated logistics provider established by the two major newspaper publishers in São Paulo, Brazil. Such collaborative logistics comprises servicing a very large number of customers in the urban setting of a large metropolitan area. The last-mile delivery is performed daily by a fleet comprising several hundreds of motorized vehicles (mainly motorcycles and light vans) that may originate either from a distribution center (that receives and consolidates different newspapers originating at the two printing sites for the delivery routes) or from informal transit points to where they are transferred, forming a multi-echelon delivery system. The resulting problem is complex not only due to its dimension but also as it involves three main types of decisions that should be taken simultaneously: the number and location of the distribution centers (DCs), the optimal synchronization of the transfer schedule of newspapers from the two printing sites to the selected DCs and the newspaper last-mile delivery. The main theoretical contribution of this work is a multi-stage solution approach whose core is a schedule-to-distribution model that allows integrating printing ramp-up constraints with the transfer shipment schedule and distribution routes. Together with a facility location model and a two-tier delivery districting, we could redesign the multi-echelon distribution operation in a more cost-efficient manner while ensuring the current service level to be maintained without service disruptions. This approach allowed us to obtain a new optimized network configuration and the corresponding delivery clusters with vehicle type choice as well as the optimal transfer schedule that yielded a significant cost reduction of nearly 18%.