The aim of this study was to classify and characterize the compositional quality of milk from river buffaloes (Bubalus bubalis) and cows (Bos spp.) in Colombia based on the fat, protein, and total solid (TS) contents. Using a hierarchical procedure, data on milk from river buffaloes (n = 7,726) and cows (n = 49,330) was filtered and subjected to cluster analysis to generate three groups: Normal (N), High quality (HQ), and Poor Quality (PQ). The categorized database was then randomly separated into two sets (training and validation) and a discriminatory analysis was applied. In total, 37.3% of river buffalo milk samples were classified as N (6.80% fat, 4.34% protein, and 16.80% TS), 13% as HQ (9.41% fat, 4.93% protein, and 19.50% TS), and 43.7% as PQ (3.95% fat, 3.92% protein, and 13.7% TS). In contrast, 41.8% of cow milk samples were classified as N (3.64% fat, 3.37% protein, 12.42% TS), 41.2% as PQ (2.71% fat, 3.08% protein, and 10.6% TS), and 16.9% as HQ (5.46% fat, 4.01% protein, and 14.82% TS). The discriminatory models for both river buffalo and cow milk were able to classify milk in the N and PQ groups with >90% accuracy and that in the HQ group with >85% accuracy.