How to prioritise leakage repairs by impact

England and Wales lose about 2,954 million litres of treated water a day through leakage. Under AMP8, the regulatory period from April 2025 to March 2030, the economic regulator Ofwat asks water companies to cut leakage by a further 457.3 Ml/d while the sector manages a £104bn investment programme.

Detection is improving, but the repair constraint remains. Acoustic loggers, satellite detection, in-pipe sensors and smart meters can all add alarms to the queue. A continuous-flow alarm, where a smart meter sees water passing for an extended period when normal use would usually stop and start, can be useful evidence. It can also be a running loo, legitimate high use or a leak outside the company's repair route.

That leaves leakage teams with a practical ranking problem: when there are more alarms than available crews, impact has to be judged through five checks that move from confidence in the alarm to the repair effort needed, then updated as better evidence comes in.

1. Confidence

Many alarms are unresolved signals when they first reach the leakage team. A continuous-flow pattern may point to a leak, but it can also come from customer behaviour or plumbing faults outside a network repair route.

The first pass should raise or lower confidence before a crew slot is committed. Meter pattern, account history, customer contact, previous visits, district metered area behaviour and nearby acoustic evidence all help separate a likely leak from a weak alarm.

Crew time is the scarce resource. A gang sent to a running loo loses a repair slot that could have cut reported leakage. Low-confidence cases can stay visible, but they should usually move into verification: customer contact, a repeat read, a boundary check or another evidence source. High-confidence alarms earn earlier field action because the visit is more likely to end in a repair.

2. Ownership

Customer-side leakage, a leak on the pipe the customer owns between the company main and the property, is commonly a quarter to a third of total leakage. When those cases sit inside the same queue as network repairs, they can quietly overload the dispatch list.

The repair route is different. A network-side leak can be planned through field operations, permits, excavation and reinstatement. A customer-side leak needs customer contact, access, ownership clarity and a different fix path. The water lost may still count against leakage performance, but the team that can close the case is different.

Ownership should be established as early as the evidence allows. Meter location, boundary checks, stop-tap information, customer-side repair history and site notes all reduce the risk of sending the wrong team. Probable customer-side leakage may deserve rapid customer contact. Confirmed network-side leakage can move into repair planning with fewer handoffs.

3. Cumulative loss

The leaks that build the largest total loss are often too small to look urgent on a single flow reading. Roughly 98% of continuous-flow leaks can sit below about 1.6 litres per minute. At that level they may be invisible to acoustic methods, but they can run for weeks.

Ranking by estimated water lost over time changes the order. A low-flow leak that has run continuously for a month can matter more than a higher-flow alarm that appeared yesterday. Flow rate still matters, but it needs duration, confidence and detection age beside it.

Smart-meter data can improve leakage work, even while coverage remains partial. Less than about 12% of meters are smart, so most leaks still sit between manual reads or other detection cycles. For the alarms that are visible, the queue should account for cumulative loss and avoid automatic preference for the highest litre-per-minute reading.

The practical test is simple: estimate how much water the leak has already lost, how much it is likely to lose before the next available repair slot, and how confident the team is that the signal is genuine.

4. Local consequence

Local consequence can move a smaller leak above a larger one. A leak near a hospital, a vulnerable customer, a watercourse, a pollution-sensitive area or a road with known service risk can justify faster action than the flow rate suggests.

The ranking should hold the site context beside the water estimate. Location can change who needs to be notified, how quickly the site should be inspected, what traffic management is needed, and whether the case creates service or environmental risk.

This part of the judgement is easy to underweight when the repair list is built from detection data alone. Sensor output rarely carries the full operational context. The stronger queue reconciles alarms with customer data, asset information, field notes, permit constraints and the known risk profile of the local network.

5. Repair effort

The same leakage saving has different value depending on what the repair takes to complete. An easy access job with clear ownership and a high chance of first-time repair can move ahead of a larger leak that needs a road closure, complex permits, specialist resource and heavy reinstatement.

The industry term for this trade-off is the Economic Level of Leakage: the point where chasing another unit of leakage costs more than the water and service benefit saved. At queue level, the question is narrower: which jobs give the best expected leakage reduction for the crew time, access effort and repair cost available this week?

Repair effort should include practical friction. Access restrictions, permit lead times, reinstatement, traffic management, crew availability, abort risk and the likelihood of a repeat visit all affect impact. A repair that looks attractive on volume can fall down the list if it absorbs scarce capacity.

A high-consequence leak or a confirmed high-loss network leak can justify an expensive repair. The queue should show the expected water saving, operational risk and repair effort beside each other before a crew slot is used.

A stale queue sends crews to the wrong job

A repair list built by hand at the start of the week can be wrong by the afternoon. New continuous-flow alarms arrive, customers respond, field teams close jobs, boundary checks change ownership, and crews find site conditions that alter the repair plan.

One case may enter the queue as probable customer-side leakage because the meter sits beyond the company main. A boundary check can move it into the network repair route. Customer contact can also remove it from the leakage queue if the flow pattern is legitimate high use. The priority should change when those facts change, without waiting for the next manual rebuild.

This becomes more important as smart-meter coverage grows. The £1.7bn rollout of 10.4 million meters is designed to close part of the visibility gap, but it also increases alarm volume. A team that already has more alarms than crews needs faster decisions as much as detection. The queue has to absorb new evidence continuously and make the next best dispatch choice.

Pressure management, mains replacement and metering all matter for AMP8. They take capital, time and programme capacity. The fastest near-term progress comes from fixing the visible leaks in the order that gives each crew slot the best expected return.