In low-resource settings, drinking-water safety decisions are frequently shaped by subjective perceptions rather than objective contamination data, yet this correspondence is poorly characterized. This cross-sectional study (n = 425) in Nadowli-Kaleo District, Upper West Region, Ghana examined concordance between perceived drinking-water quality and field-observed sanitary contamination risk, together with socioeconomic determinants of risk appraisal calibration. Perceived quality was rated on a 1–5 Likert scale; field-observed sanitary contamination risk was aggregated into a 0–4 ordinal risk index from a structured observation checklist (a proxy for, not a direct microbiological measure of, water quality), reverse-scaled, with perceptions classified as calibrated when |gap| ≤ 0.5. Perceived quality was strongly and inversely associated with observed contamination risk (Spearman’s ρ = −0.780; p < 0.001), indicating that household perceptions tracked broad contamination-risk gradients well. However, agreement at the individual household level was weaker: only 49.4% of households met the calibration criterion (mean gap = −0.86) under our primary threshold, and this proportion was sensitive to the threshold chosen (ranging from 26.1% at ±0.3 to 63.1% at ±1.0). Educational attainment showed no reliable association with calibration accuracy (no formal education 55.5%; secondary 44.1%; postgraduate 12.5%, n = 8; Pearson χ²(4) = 7.24, p = 0.124), and a quadratic logistic model provided no evidence of a non-linear (inverted-U) relationship (quadratic OR = 0.94; p = 0.462); income and residence duration were likewise not significant predictors. These findings suggest that perception-based metrics alone may be insufficient for individual-level risk assessment, and that common socioeconomic proxies do not reliably identify well-calibrated households, calling for objective monitoring and tailored risk communication to advance SDG 6.