In the supply chain of Fast-Moving Consumer Goods (FMCG), logistics costs represent a major part of total expenses. At these low-level chains, one usually faces a Vehicle Routing Problem (VRP). In practice, however, due to the high cost of service in many cases, some customers are not selected to serve. Investment-related restrictions in many cases make it impossible to serve some of the potential customers. In such conditions, designing a supply chain network, including a location-allocation problem in the warehouse, Multiple Depot Vehicle Routing Problem (MDVRP) at the distribution level, and customer selection at the retail level in several periods of time, is considered. In this respect, in addition to certain methods that can be used in small sizes, metaheuristic algorithms have been used to solve large-scale models. With the aim of improving the performance, if not improving a few diversi cations, algorithms are temporarily enhanced; eventually, by using statistical approaches, it has been demonstrated that this method could have a signi cant impact on the quality of responses. Genetic Algorithm (GA) and Simulated Annealing (SA) algorithm have been used for this purpose.