This study addresses the Vehicle Routing Problem with Time Windows (VRPTW) in the context of ice distribution by introducing a novel mathematical model that incorporates practical constraints essential for real-world applications. These constraints include customer retention strategies and quality preservation methods, which are important for maintaining customer satisfaction and product integrity. The objective is to minimize the total costs, including fuel expenses, standard and bonus driver wages, missed delivery penalties, and costs related to a quality preservation strategy. Given the NP-hard nature of this problem, this study proposes a hierarchical cluster-first-route-second approach and a Differential Evolution (DE) algorithm to solve large-scale problems. The effectiveness of these methods was examined and compared through test cases involving various problem sizes using real-world data from an ice distribution company in Thailand. The results show that the hierarchical cluster-first-route-second approach is more effective for the practical problem. Using capacitated K-means clustering, this hierarchical approach groups customers, enabling the solution of manageable subproblems through Mixed-Integer Linear Programming (MILP). The proposed method not only provides cost-effective and scalable solutions, but also outperforms traditional methods in terms of computation time and feasibility for large-scale applications. This study offers significant theoretical contributions by extending VRPTW models and providing practical implications for optimizing distribution strategies in competitive market environments, leading to substantial cost reductions and enhanced operational efficiency.