This study develops a decision-support framework for a real-world multi-depot distribution planning problem observed at a leading large-scale supermarket chain in Turkey. The operational setting involves two distribution centers, geographically dispersed retail stores, multiple product groups, pallet-based demand, heterogeneous vehicles, depot throughput limits, product-vehicle compatibility rules, time windows, and distance-dependent transportation costs. A mixed integer linear programming (MILP) model is first formulated to represent the integrated depot assignment, product allocation, vehicle activation, and routing decisions. Since the full model becomes computationally challenging as the number of stores increases, an Adaptive Large Neighborhood Search (ALNS)-based matheuristic is designed to explore assignment decisions while using exact optimization-based routing evaluations for candidate solutions. Computational experiments compare the ALNS framework with time-limited MILP runs on representative instances. The results indicate that the proposed method can produce feasible and competitive solutions within shorter computational times than the monolithic MILP on the tested instances. The study therefore positions the ALNS framework as a scalable analytical tool for tactical distribution planning rather than as an empirically validated source of cost savings against current operational practices, for which historical baseline data were not available.

Keywords

Retail logistics; Multi-depot vehicle routing; Adaptive large neighborhood search; Mixed integer linear programming; Heterogeneous fleet optimization