A simulation–optimisation framework for front-warehouse location planning in instant retail

The instant retail model that provides ultra-fast on-demand delivery service from local front warehouses is gaining increasing popularity. However, its profitability is constrained by escalating costs due to system characteristics like temporal and spatial concentration of demand, time-varying courier availability, interdependence among consecutive orders, and the consequent cascading order delays. This paper develops a simulation-optimisation framework to solve the front-warehouse location problem in instant retail, which integrates a discrete-event simulation model that captures system characteristics and micro-level system dynamics, and proposes a ranking-and-selection algorithm that identifies the best location strategy with high statistical confidence via adaptive sampling. Extensive numerical studies using both synthetic and real data demonstrate that this approach consistently selects the optimal solution with over 99% probability while maintaining high computational efficiency even for large-scale problems. This approach outperforms classical facility location models and simulation-based genetic algorithms in terms of both profitability and service timeliness, and the advantage is primarily driven by the feature of temporal demand concentration arising in instant retail. We further find that under constrained courier capacity, light order batching can effectively improve profits and service responsiveness, and off-peak pricing may mitigate cascading order delays when demand peaks coincide with traffic congestion.