In this work, dwell time refers to the period that a container spends in drayage, encompassing the duration from when it is picked up from the port or terminal until it is delivered to its final destination. Accurate estimations of these times enhance operational efficiency by enabling better resource allocation, such as labor, equipment, and space, and allowing for precise scheduling of container movements. This reduces idle times, increases throughput, and lowers storage costs and handling fees. Accurate predictions also help avoiding detention charges, late delivery fees, and other potential penalties. However, the data used for such estimations have several difficulties, such as of many occurrences with similar features but different dwell times, creating misleading values when using aggregation functions or higher errors when using standard approaches of machine learning prediction techniques. A new approach is proposed to minimize such issues, enabling the determination of accurate estimations and enriching the information given to users. By understanding the characteristics of different operational distribution clusters, logistics operators can optimize resource allocation, streamline processes, and enhance overall efficiency, reducing costs, faster turnaround times, and increasing customer satisfaction. A case study of the Port of Sines is considered, demonstrating the potential of the proposed approach to improve drayage operations and contribute to a more efficient and resilient supply chain.