Estimating the distribution of domestic water consumption among households is important for establishing baselines against which policy success can be measured. Estimates of this kind are typically produced using household surveys. This study aimed to investigate how simple heuristic estimates of water consumption and surveys compare with measures of water consumption obtained from meter readings. It also aimed to characterise (parameterise) the distribution of metered water consumption in Canberra. The ‘gold standard’ measure of consumption was 7 years of quarterly metered readings of 27,174 dwellings in Canberra. Heuristic estimates and survey-derived estimates were compared with these actual meter readings. Survey estimates were obtained by population weighting an online panel survey that asked respondents about their consumption behaviours. Heuristic estimates were obtained on the assumption that total water consumption can be approximated by combining a priori estimates of consumption for toilet use, shower use, washing of dishes, washing machine use, teeth brushing, cooking and drinking. Heuristic estimates of consumption underestimated extreme values. Survey derived estimates provided unreliable indicators of actual consumption. End-use analysis typically involves expensive and detailed house visit surveys or participant reporting surveys. This study suggests that both self-report surveys and heuristic models should be treated with caution. The extreme variation in distribution of consumption among households requires population sampling much larger than normally undertaken. Nine common mathematical distributions were parameterised using the metered household consumption. This work has application in evaluating policies aiming to modify water consumption.