Client
A UK regional water and wastewater utility serving several million household and business customers, entering the AMP8 regulatory period with tightened leakage and asset health reporting obligations
Sector
Energy & Utilities
Engagement
Consolidated leakage detection, network telemetry and asset condition data platform, plus a prioritisation model for mains replacement and repair - multi-quarter programme.
What the client needed
Like most UK water utilities, our client's network telemetry had grown up asset by asset over decades: acoustic loggers from one supplier, pressure sensors from another, a SCADA historian nobody fully trusted, and a leakage team that ultimately worked from spreadsheets stitched together from exports out of each. Ofwat's PR24 price review set tougher leakage reduction and asset health reporting expectations for the AMP8 period running to 2030, with real financial consequences attached to underperformance, and the client's board wanted a credible plan for meeting them before the regulator asked the question directly. The deeper problem wasn't a lack of data. Sensor coverage across the network was actually reasonable. The problem was that no single system could tell an engineer, in one place, which section of ageing main combined the highest leak probability with the highest consequence of failure, so prioritisation decisions were made on institutional memory and whichever spreadsheet was most recently updated, not a consistent model applied network-wide.
How we worked
- Built ingestion pipelines from every telemetry source - acoustic loggers, pressure and flow sensors, the SCADA historian and manual inspection records - into a single network data model keyed to a common asset register.
- Reconciled the asset register itself first, since duplicate and mismatched asset IDs across legacy systems were quietly corrupting every downstream analysis before we started.
- Built a leak-likelihood and consequence-of-failure scoring model per network section, combining sensor anomaly signals, pipe age and material, burst history and proximity to critical customers or infrastructure.
- Replaced the leakage team's manual spreadsheet workflow with a prioritised, continuously updated work list, ranked by combined risk score rather than whichever area last generated a customer complaint.
- Built regulatory reporting views mapped directly to Ofwat's AMP8 leakage and asset health metrics, so reporting became a query against live data rather than a quarterly manual compilation exercise.
- Ran a phased rollout by region, validating the scoring model's recommendations against field crew ground truth before extending it network-wide.
Measured results
All figures verified with the client. Specific site and network detail withheld in line with our standard confidentiality terms and regulatory sensitivity.
- The leakage team now works from a single prioritised list instead of reconciling exports from five separate systems, cutting the time spent compiling a weekly work plan from most of a day to under an hour.
- Field crews report the risk-scored work list matches or improves on experienced engineers' own judgement in the large majority of cases, and is now the default starting point rather than a secondary reference.
- Quarterly regulatory reporting against AMP8 leakage and asset metrics moved from a multi-week manual compilation exercise to a same-day process against live data.
- Reconciling the asset register surfaced a meaningful number of duplicate and orphaned asset records that had been silently skewing prior leakage estimates, improving the accuracy of the client's own baseline figures.
- The prioritisation model is now being extended to inform the client's next AMP investment planning cycle, beyond its original leak-detection scope.
"We had more sensor data than we knew what to do with and still couldn't answer 'which pipe first' with any confidence. The platform didn't add new sensors, it made the ones we already had trustworthy enough to build a work plan around."
Working on something similar?
If this engagement looks like the kind of problem you are facing, we would be glad to compare notes by email.
Context and constraints
Water utilities sit under a form of regulatory scrutiny that few other infrastructure sectors match. Ofwat's PR24 price review, covering the AMP8 investment period to 2030, tightened leakage reduction targets and attached real financial mechanisms to performance against them, arriving alongside heightened public and media attention on network resilience more broadly. Our client's leadership was clear from the outset that this engagement wasn't a data modernisation project that happened to touch leakage, it was a leakage and asset performance problem that happened to require fixing the data first. That ordering mattered for scope discipline throughout: every technical decision was tested against whether it moved the leakage and asset-health numbers the regulator would actually assess, not against what looked most sophisticated on a platform roadmap.
The asset register problem took longer to fix than anyone initially budgeted for, and we pushed to fix it first rather than build analytics on top of data we didn't yet trust. Duplicate and mismatched asset identifiers between the SCADA historian, the GIS system and the maintenance records meant the same physical section of main could appear under three different IDs with three different histories, and any scoring model built on top of that would have produced confident, precise, and wrong prioritisation. Reconciling the register was unglamorous work with no visible output for several months, and it was also the single highest-leverage thing we did on the programme.
Building a scoring model field engineers would actually trust
A risk score that field crews ignore in favour of their own judgement isn't a useful output, so we validated the leak-likelihood and consequence-of-failure model against experienced engineers' own prioritisation calls before asking anyone to rely on it operationally. Where the model and an engineer's instinct disagreed, we treated that as a signal to investigate rather than a sign the model needed to be forced into agreement, and several of those disagreements led to genuine improvements in the scoring logic, most usefully around how burst history should be weighted for older cast iron mains versus newer polyethylene sections. That validation loop, run region by region during the phased rollout, is the main reason field adoption held up once the tool moved from pilot to default.
Reporting as a byproduct, not a separate project
Because the regulatory reporting views were built directly against the same live data model as the operational scoring tool, quarterly AMP8 reporting stopped being a distinct, dreaded compilation exercise and became a query against data the leakage team was already using day to day. That single design decision, reporting and operations sharing one source of truth rather than a duplicated reporting data mart, removed an entire category of the reconciliation errors and late-quarter scrambles that had characterised the client's previous reporting cycle.
Lessons learned
The first lesson was that in a heavily instrumented network, the constraint is rarely sensor coverage, it's whether anyone can trust and combine what the sensors are already reporting. Adding more monitoring onto an unreconciled asset register would only have produced more precisely wrong answers, faster.
The second lesson was that a prioritisation model earns operational trust in the field, not in a steering committee. Validating against experienced engineers' judgement, and being willing to change the model when they disagreed with it, did more for adoption than any amount of executive sponsorship could have.
The third lesson was that building regulatory reporting on the same live data model as day-to-day operations turns compliance from a recurring cost centre into a natural byproduct of doing the operational work well, which is a far more durable position heading into the next price review than a reporting process that only exists to satisfy the current one.
If your organisation is facing a similar combination of tightening regulatory reporting requirements and network data that's more fragmented than it first appears, we would be glad to discuss what a programme like this might look like for you. Email sales@halfteck.com.