In the process of forming clusters with the K-Medoids algorithm, cluster result anomalies often occur, such as outliers. This value appears as a revelation in existing data patterns. Outliers occur due to measurement errors, rare events, or due to other unexpected factors. In this research, the dataset used is data on prospective KIP recipient students at Budi Darma University, where there is a high level of interest in KIP Kuliah while the quota is limited, which means that KIP Kuliah administrators sometimes have difficulty determining which students are eligible to receive KIP Kuliah. For this reason, the K-Medoids clustering technique was used to cluster data on 54 prospective students who were eligible to receive KIP Kuliah Merdeka and those who were not eligible. From the cluster results, outlier detection was carried out using the box plot method with the aim of finding out whether each cluster member was actually in the appropriate cluster or not. The result is that the data cluster is divided into 2 (K-2). In the max min centroid selection, cluster I consists of 52 members and cluster II consists of 2 members, where the outlier data consists of 3 data, while in random centroid selection (python), cluster I consists of 36 members and cluster II 18 members with data The outlier consists of 4 members. The accuracy of the clustering results between max min and random centroid selection has an accuracy of 64.81%, and the outlier accuracy is 75%.