Sub-Theme Water Sustainability (PAIP)

Non-Revenue Water (NRW) is water that a utility treats and puts into its distribution network but never bills: lost to pipe leakage, meter under-registration, unauthorised connections, and reservoir overflow. It is the central efficiency measure in water utility management: every cubic metre lost has already absorbed the full cost of abstraction, treatment, chemicals, and pumping.

Malaysia's national NRW rate sits around 35%, against a policy target closer to 25%. Pengurusan Air Pahang Berhad (PAIP) operates 74 treatment plants across all 11 Pahang districts and publishes monthly production, billed consumption, and NRW figures for each.

Every water utility faces the same constraint: it can see how much water it is losing, but it always has more leaking network than it has crews, budget, and time to repair. The operational question is not how large the losses are, which PAIP already knows, but which plants to fix first.

The difficulty is that the industry-standard measure answers a different question. NRW is reported as a percentage because a percentage is what makes plants comparable to one another and to national targets. But a percentage describes how efficient a plant is, not how much water is actually recoverable there. A small plant losing half of a small output posts an alarming rate; a large plant losing a fifth of a large output posts an acceptable one, while losing many times more water. Ranking by rate and ranking by volume therefore produce substantially different repair queues, and only one of them maximises water saved per crew-day.

The losses are also worth pursuing. Monthly NRW is flat across 2025: the median sits between 48.9% and 55.6% in every month, with no wet-season or dry-season signal, which points to continuous physical leakage rather than seasonal demand swings or intermittent meter and billing errors. Losses of that kind do not correct themselves, but they are precisely the kind that repair work recovers.

The problem is therefore that PAIP already publishes the data needed to plan repairs, but in a unit built for performance reporting rather than for targeting. No tool currently converts those monthly figures into a ranked, volume-weighted intervention schedule.

1. To measure water loss at PAIP plants two ways, as a percentage of what each plant produces and as an absolute volume, and show how differently the two measures order the plants.
2. To deliver a dashboard showing both measures side by side (percentage for judging how a plant is performing, volume for deciding where to send repair crews) so the difference cannot be overlooked.
3. To build a Leakage Intervention Priority Score (LIPS) that ranks plants by how much water a repair would actually recover, rather than by loss percentage.

• A cleaned, tidy, long-format dataset of 888 plant-month records with a documented plant-name crosswalk; the published workbook is not analysis-ready in its current form.
• Quantified evidence that rate-based and volume-based prioritisation produce materially different intervention orders.
• A ranked LIPS intervention schedule for all plants.
• A dashboard presenting both metrics side by side.

• The dataset covers a single year (2025), so trends cannot be distinguished from year-specific events and the anomaly detector has only twelve observations per plant.
• No cost data is available. Volume lost is not the same as volume economically recoverable: a plant losing 40 million m³ through a dispersed network of ageing pipes may be far more expensive to fix per m³ than one losing 2 million through a single identifiable main.
• NRW does not distinguish real losses (physical leakage) from apparent losses (meter error, theft, billing lag). These require entirely different interventions. Without a formal water balance breakdown, the analysis should present combined losses and be explicit that it cannot separate them.
• Negative NRW records point to measurement or timing problems that this dataset cannot diagnose; they should be flagged for PAIP rather than silently corrected or deleted.
• Tariff figures are not included, so financial loss estimates should not be presented unless PAIP's published tariff schedule is sourced and cited.