Water losses in drinking water networks are diagnosed from uncertain measurements, especially where instrumentation is scarce. Choosing wrongly between reducing real losses (leaks) and apparent losses (metering errors and unauthorised consumption) wastes investment, a concern in Morocco. This study combines the standard water balance of the International Water Association with a Monte Carlo framework that expresses the consequences of measurement uncertainty as three decision indicators: the probability of misclassifying the dominant loss component, the ranking stability of intervention options, and the regret, in cost per cubic metre saved, of a wrong choice. It was applied to four virtual sectors representing contrasting Moroccan urban networks at four levels of instrumentation. Their parameters come from published ranges and assumptions, without field data, so the probabilities describe these sectors and not real networks. From the lowest to the highest level, the probability of misclassification fell from 0.002–0.12 to below 10^-9 in the three sectors where real losses dominate, and from 0.36 to 0.05 where apparent losses dominate, a gain only with the audit of unauthorised consumption and billing data. Ranking stability (Kendall's coefficient) rose from 0.14–0.58 to 0.53–0.98, and the regret fell by more than half. A 5% bias of the inlet meter, the most influential parameter, raised the misclassification probability in the sector with dominant apparent losses to 0.84 at the highest level. On these sectors, the framework identifies where to measure first before investing in leak detection or meter replacement.