Non-revenue water is a major pressure on drinking-water resources in semi-arid regions, where every cubic metre lost has already been abstracted, treated and pumped. This study evaluates a two-year active leakage control programme carried out by the national utility in Aïn Aouda, a Moroccan town of 24 000 service connections, between May 2024 and April 2026. A field register of 1 229 interventions (1 187 leak repairs, 92% georeferenced) was coded by leak type, failed component, detection mode and service impact, and combined with the survey effort of each campaign (2 070 km of mains surveyed), ten bimonthly minimum night flow campaigns recorded over three consecutive nights with calibrated ultrasonic flow meters under a fixed supply configuration, eight pressure-logging campaigns and continuous flow measurements at the reservoirs. Trends were tested with autocorrelation-corrected Mann-Kendall tests and negative binomial regression, and spatial patterns with kernel density estimation, Ripley's L function, Moran's I and Getis-Ord statistics. Service connections accounted for 60% of repairs, and 66% of them were hidden leaks found only by acoustic survey, against 39% for mains; three components explained 84% of failures. Repairs decreased significantly (−6.7% per month, 95% confidence interval −9.4 to −3.9%) while the minimum night flow fell from 274 to 154 m³/h in six months of intensive search, rose to 189 m³/h when search intensity halved, and reached 127 m³/h (−54%) after a renewed campaign; hidden leaks found per campaign followed the length of mains surveyed (ρ = 0.74). With a measured night-day factor of 20–21 h, real losses are estimated to have decreased from 47% (42–54%) to 19% (16–23%) of the distribution input, an estimated saving of 1.1 million m³ (1.0–1.2), 0.65–0.82 GWh and 390–550 t CO₂ per year (model-based estimates with propagated uncertainty). Leak density concentrated in three persistent hotspots (11% of the urban area holds 67% of repairs, p < 0.01), and main failures caused 82% of the 22 494 customer-hours of supply interruption. The results document the response of night flow to a sustained active leakage control effort and its rebound when effort declines, and provide a spatial basis for district metering, pipe renewal and pressure management.