Clean water is a basic need that is very important for human life. In Indonesia, many clean water companies supply water for community needs. Indonesia's government-owned clean water company has many customers spread across various provinces. This research aims to apply the k-medoids algorithm in clustering PAM customers based on provinces in Indonesia to determine which provinces are included in the low, medium, and high-level clusters of PAM customers. The data used in this research is on the number of PAM customers, the amount of water distributed, and the value of the water distributed from 1998 to 2021 to 34 provinces in Indonesia. This research obtained 3 clusters, namely low, medium, and high-level clusters determined by the number of PAM customers, the amount of water distributed, and the value of the water distributed in each province. Cluster 1 (C1) is a low-level cluster that obtained results from 19 provinces, Cluster 2 (C2) is a medium-level cluster that obtained results from 11 provinces, and Cluster 3 (C3) is a high-level cluster that obtained results from 4 provinces. And get a k-medoids cluster validation value of 0.014846718. In this way, it is hoped that the government can identify areas in each province that require improvement or improvement in making appropriate decisions regarding clean water resources to provide better services to customers