Designing parcel delivery areas under customer preferences for self-collection: A probabilistic districting approach

Motivated by the growth of e-commerce and the widespread adoption of self-collection strategies, this paper addresses a districting problem arising in last-mile logistics. The objective is to divide an urban area into a given number of compact and contiguous districts—-each assigned to a specific driver for deliveries—-that also satisfy maximum (expected) workload constraints. End-users’ preferences for alternative delivery options (home delivery vs. self-collection) are modeled using a Multinomial Logit rule. These preferences are then used to estimate the daily visit probabilities for both pick-up points and customer nodes. The resulting uncertainty in customers’ presence and demand motivates the development of a simheuristic solution procedure, which optimizes district design while employing a simulation routine to estimate the workload of each district—-and, by extension, each driver. Leveraging real-world data from an Italian logistics provider, the proposed approach enables the enhancement of delivery operations based on optimized pick-up point location scenarios and the quantification of the associated (proxied) economic impacts. Ultimately, it demonstrates how self-collection leads to robust solutions that hedge reasonably well against varying operational conditions and projected demand volumes compared to traditional home delivery.