Accurate water-quality assessment is vital, but laboratory costs limit monitoring in many regions. We test whether a small, low-cost indicator panel can classify Water Quality Index (WQI) categories in Nuevo León, Mexico. Using 1,302 REMANECA samples, we computed WQI with a weighted multiplicative model and trained five classifiers (RF, SVM, DT, KNN, NB) on physicochemical features. Cross-validation ranked Random Forest (RF) best with 11 indicators (accuracy \(0.921\pm 0.023\) ; weighted F1 \(0.912\pm 0.028\) ; macro precision \(0.926\pm 0.037\) ; macro recall \(0.785\pm 0.073\) ). Feature selection and importances emphasized total hardness, coliforms, nutrients (PO \(_4\) , NO \(_3^{-}\) , NH \(_3\) ), and pH. A cost-aware five-test panel (hardness, PO \(_4\) , pH, NH \(_3\) , SST) retained strong performance (RF accuracy \(0.857\pm 0.026\) ; weighted F1 \(0.829\pm 0.030\) ) with reduced minority-class sensitivity (macro recall \(0.615\pm 0.059\) ). Errors concentrated between adjacent categories; detection of heavily contaminated water remained stable (recall 98% to 97%) and the majority class stayed high (99% to 98%), while “excellent” and “slightly contaminated” degraded. These results show that reliable WQI classification is achievable with a compact, low-cost indicator set. A tiered strategy—screen with the five-test panel and confirm with the full suite—can expand coverage under fixed budgets while preserving identification of severe contamination.