The IWA water balance explained: every component, with a worked example

The IWA standard water balance is a table with four columns. Reading left to right, it splits the water put into a system into smaller and smaller pieces until every cubic metre is either billed, legitimately used, lost through leaks, or lost through metering and billing failures. It was published by the IWA Task Force on Water Losses in 2000 and has since become the basis of the AWWA M36 water audit, the World Bank's NRW guidance and most national regulatory reporting.

The four columns

Column 1 is what went in. Column 2 asks whether it was used legitimately or lost. Column 3 asks whether the legitimate use was billed, and whether the loss was physical (real) or commercial (apparent). Column 4 answers the money question: did the utility get paid for it?

System input volume (SIV)

The volume of treated water entering the distribution system over the balance period, from every source: treatment plants, bores, springs, bulk imports. It is measured by bulk meters, which are the most accurate numbers in the balance — typically ±1–3% if calibrated. Water exported to another utility is normally shown as a billed (or unbilled) bulk item so it can be excluded from customer-meter error calculations.

Authorised consumption

Water used by registered customers and by the utility itself for legitimate purposes. It splits four ways: billed metered (the billing system's metered volume, adjusted to the period), billed unmetered (flat-rate or assessed customers), unbilled metered (metered but not charged — a council depot, a school on an exemption, an unbilled bulk transfer) and unbilled unmetered (firefighting, mains flushing, street cleaning, works).

Water losses = SIV − authorised consumption

Everything else. Water losses are the sum of apparent losses and real losses, and this is where the balance stops being accounting and starts being estimation.

Apparent losses

Water that reached a customer but was never recorded: unauthorised consumption (illegal connections, bypasses, tampered meters, hydrant theft), customer meter under-registration (meters that slow with age and miss low flows), and data-handling errors (accounts missing from billing, estimated reads never trued up). Each is estimated from surveys, meter tests and billing audits. Because every litre would have been billed, apparent losses are valued at the full tariff.

Real losses = water losses − apparent losses

Physical leakage from mains, service connections and storage, up to the customer meter. In a top-down balance real losses are the residual, which means they inherit every uncertainty above them. That is why a good balance always reports a ± band on real losses — often ±10–30% — and why the figure should be cross-checked bottom-up with night-flow measurement in district metered areas.

Revenue water and non-revenue water

Worked example

A utility puts 22.7 million m³ into supply in a year. Billed metered consumption is 15.7 million m³ and billed unmetered 0.35 million m³, so billed authorised consumption is 16.05 million m³ and NRW is 6.65 million m³ (29.3%). Unbilled authorised use is 0.21 million m³. Apparent losses are estimated at 0.87 million m³ (±24%): 0.2 million from an estimated 600 illegal connections and 300 tampered meters, 0.65 million from 4% meter under-registration, and a small data-handling allowance. Real losses are therefore 22.7 − 16.26 − 0.87 = 5.57 million m³ (±10%), or 15,260 m³/day. With 1,800 km of mains, 86,000 connections and 43.5 m average pressure, UARL is 4,970 m³/day and the ILI is 3.07 — band B.

Common mistakes

• Using calendar-year billing without correcting for meter-reading lag. Quarterly reads mean up to three months of consumption belong to a different year.
• Counting exported water as consumption of your own customers, which inflates the meter-error estimate.
• Treating NRW % as a leakage indicator. It is a financial indicator; use litres per connection per day and the ILI for leakage.
• Forgetting inactive accounts with a live service pipe — they still leak and belong in the connection count.
• Entering optimistic uncertainties. ±2% on a sample-based estimate of illegal use is not credible; ±50% is.