The rapid growth in e-commerce volumes, coupled with customer expectations of faster, flexible and cheaper parcel deliveries is increasing the pressure on retailers to design the most efficient delivery network. Collection-and-delivery points (CDPs aggregate demand for both forward and return flows and help minimize additional tours in the case of failed deliveries. In this paper, we formulate an optimization model that integrates CDPs in the design of the overall distribution network, including the location of upstream transshipment facilities. The model accounts for changes in demand density due to the placement of CDPs. It considers demand aggregation at the CDP for both forward and return flows, and the impact of failed deliveries and failed return pickups on the routing cost. The model considers different route options and solves them using extended routing cost approximation formulae thus allowing the implementation of the model on large-scale problems. We then apply the model to solve a real-world case study on the distribution network of a major Brazilian retailer. The results demonstrate that failed deliveries and failed return pickups increase both the last-mile cost and the overall cost of distribution, and CDPs effectively reduce these costs by aggregating the demand and minimizing travel time.

电子商务量的快速增长，加上客户对更快、更灵活和更便宜的包裹递送的期望，正在增加零售商设计最高效递送网络的压力。收集和交付点（CDP 聚合了前向和回流的需求，并有助于在交付失败的情况下最大程度地减少额外的旅行。在本文中，我们制定了一个优化模型，将 CDP 集成到整个分销网络的设计中，包括上游转运设施的位置。该模型考虑了由于放置 CDP 引起的需求密度变化。它考虑了在 CDP 处对前向和回流的需求聚合，以及失败的交付和失败的回程取货对路线成本的影响. 该模型考虑了不同的路线选项，并使用扩展的路由成本近似公式来解决它们，从而允许在大规模问题上实施该模型。然后，我们应用该模型来解决有关巴西一家主要零售商的分销网络的真实案例研究。结果表明，失败的交付和失败的回程提货会增加最后一英里的成本和总配送成本，而 CDP 通过聚合需求和最小化旅行时间有效地降低了这些成本。