In order to find coherent clusters of facilities based on geographic coordinate, user/inspector ratings, and the range of services provided, this study applies unsupervised machine learning to 6,520 health facility records from throughout Uganda. Three stable clusters that correspond with broad regional divisions in Uganda were produced by K Means following systematic cleaning, robust imputation of missing ratings, removal of obvious geographic outliers, feature scaling, and objective selection of the number of clusters using the Silhouette method: a Central/Eastern cluster with high facility density and diversity of services, a Northern cluster with a higher proportion of government-operated facilities and fewer specialized services, and a Southern/Western cluster with more balanced ownership and a relative strength in maternity and laboratory services. We also used centroid analysis, a multilabel binarization procedure to extract the prevalence of critical services, and the distribution of care systems (GOVT, PNFP, and PFP) to further characterize the clusters. The clusters’ remarkable correlation to established regional borders was validated spatially using an administrative district shapefile. The findings are examined in light of Uganda’s health planning requirements, and suggestions for focused action, more data gathering, and additional modelling are made.