The vehicle routing problem (VRP) under study involves managing a fleet of trucks existing to provide transportation services from a centralized distribution center (DC) to a set of geographically dispersed customers. The problem is motivated by a real problem from one of the large chains of high-end retail stores in Thailand. The objective of this study is to develop a simple and effective algorithm that can help the user manage the fleet in order to reduce the total transportation cost. The developed algorithm is based on the parallel savings algorithm but with a stochastic component that has a capability of escaping the local optima. The stochastic component applied a biased random sampling (BRS) and regret based random sampling (RBRS) to the selection of customers to add to delivery routes. Specifically, the selection probabilities are computed proportionally from the savings value of merging customers. The algorithm generates solutions containing delivery routes and schedules for all trucks in the fleet that satisfy daily store demands under limited truck capacity and delivery time window restrictions. Three measures of performance include the total distance, total travel time, and total cost. Computational test of the algorithm was performed on a set of standard problems, and promising results are obtained.