Water losses are critical indicators of operational and commercial efficiency in water supply systems. This study investigates the determinants of water distribution losses in the state of Santa Catarina, Brazil, using official data from the National Sanitation Information System (SNIS) for the year 2022, encompassing 293 municipalities and a population of approximately 7.6 million. A three-phase analytical approach was adopted: exploratory analysis and dimensionality reduction through Principal Component Analysis (PCA); classification of municipalities using the CHAID decision tree algorithm; and interpretation of the resulting predictive model. The results identified 28 key variables explaining 75% of the variance in the dataset. Water consumption index, revenue loss, macro-metering, and water export volumes emerged as significant predictors of loss rates. The study found an inverse correlation between water consumption and loss rates - municipalities with consumption indices below 55.8% exhibited average losses of 52.5%, while those above 68.6% had losses reduced to 21.4%. The findings emphasize the need for integrated metering, infrastructure optimization, and tailored strategies based on municipal cluster profiles to reduce water loss and improve system performance. This methodology provides a replicable framework for data-driven decision-making in public water management.