How to calculate ILI, step by step.

The arithmetic at the end of an ILI calculation is one division. The work is everything that produces the two numbers going into it — and most of that work is in the numerator, because real losses are not measured directly. They are what is left after you have accounted for everything else.

This walks the whole calculation through on a single network.

Before you start: fix the period and the boundary

Two decisions that are easy to skip and expensive to revisit.

The period. Use a full year for an annual ILI assessment and align all volumes to the same dates. Shorter periods can help with operational monitoring, but seasonal demand and repairs affect comparisons. This example assumes 365 days of continuous supply.

The boundary. Identify every inlet and outlet of the network being assessed. Include imported water in system input and record exported water as authorised consumption. If you instead calculate a distribution-only balance after subtracting exports from input, exclude those exports from consumption too. Never subtract them twice. Our example has no exports.

Step 1: Build the water balance down to real losses

Use the IWA water balance to estimate current annual real losses (CARL). Two subtractions give the result:

Water losses = System input volume − Authorised consumption

Real losses (CARL) = Water losses − Apparent losses

Our example network for the year:

280,000 m³ a year is 767 m³/day. That is the numerator.

Three notes on the rows people get wrong:

• Unbilled unmetered consumption is not leakage, but it only stays out of your leakage figure if somebody writes it down. Hydrant flushing, mains disinfection, firefighting, street cleaning. Most utilities estimate this line; estimating it low inflates your real losses.
• Customer meter errors contribute to apparent losses. Use test data or a documented estimate. If meters under-record consumption and you do not account for it, the balance overstates real losses. Meter age alone does not tell you the size of the error.
• Data-handling error is real. Estimated reads never reconciled, accounts missing from the export, unit conversion mistakes. It belongs on the apparent side.

Step 2: Collect the four UARL inputs

UARL means unavoidable annual real losses: a modelled reference level for a well-managed network. The formula below expresses it as a daily volume. Use the same units for CARL and UARL before dividing.

UARL (litres/day) = (18 × Lm + 0.8 × Nc + 25 × Lp) × P

Each input has a usual source and a usual failure:

Lp is the total length of underground pipe from property boundaries to customer meters across the network. Estimate it as connection count × average underground length, then convert metres to kilometres. In this example, 7,500 × 2 m = 15 km. Use zero only where meters really are at the boundary.

Pressure multiplies the whole UARL expression. With CARL unchanged, overestimating pressure raises UARL and lowers ILI. Use a representative average for the network and assessment period.

Step 3: Compute UARL

For our network:

UARL = (18 × 210 + 0.8 × 7,500 + 25 × 15) × 42

UARL = (3,780 + 6,000 + 375) × 42

UARL = 10,155 × 42 = 426,510 litres/day, or 426.5 m³/day

Step 4: Divide

ILI = CARL / UARL = 767 / 426.5 = 1.80

Step 5: Check the result

Four checks before you use the result:

1. Is ILI below 1.0? Check the water balance and whether the UARL model fits the network. A low result can indicate input errors, but is not automatically impossible: small networks at low pressures can fall below the standard estimate. See the UARL model limitations.
2. Is ILI unexpectedly high? Check units, mains length, connection count and pressure before interpreting it as poor performance.
3. Do losses per connection look plausible? Here, 767,000 litres divided by 7,500 connections is about 102 litres per connection per day. A large difference from comparable networks warrants investigation.
4. Does a night-flow estimate support the balance? Subtract legitimate night use from minimum night flow, then allow for pressure changes through the day before estimating daily leakage. Compare the same units and period; a night flow in m³/hour cannot be compared directly with annual losses.

The percentage trap, in one example

Non-revenue water is system input minus billed authorised consumption: 1,500,000 − 1,160,000 = 340,000 m³, or 22.7% of system input. This includes 20,000 m³ of unbilled authorised consumption, 40,000 m³ of apparent losses and 280,000 m³ of real losses. It is not all leakage.

The same network has an ILI of 1.80: estimated real losses are about 1.8 times the UARL benchmark for its size and pressure. This alone does not establish whether further leakage reduction is worthwhile.

The two indicators answer different questions. Non-revenue water describes the share of input that is not billed. ILI compares real losses with a network-specific benchmark. Lower consumption can increase the percentage even when leakage is unchanged, so report both with their context.

Use the percentage to explain the water balance and ILI to support a technical comparison of leakage.

Step 6: Write down how you got there

Whatever produced the number has to survive being asked about — by a regulator, an auditor, or your own successor. Keep the assessment period and boundary definition, every balance component with its source, an explicit note on which components were estimated and how, the four UARL inputs with their dates, and the pressure measurement method.

A number nobody can reconstruct is a number that has to be redone.

Where the platform fits

The balance is the hard part, and it is hard mostly because of timing and data quality rather than arithmetic. Water loss management on Divako keeps system input, billed consumption and the zone hierarchy in one place, validates readings before they are subtracted, and recomputes the balance as data lands rather than once a year — so the CARL estimate can be traced back to its inputs and updated as better data becomes available.

You still need asset records, representative pressure measurements and documented estimates for unmetered consumption and apparent losses. The platform helps organise the evidence; the calculation remains only as reliable as its inputs.

For the regulatory context — who has to assess, by when, and what happens in 2028 — see the Drinking Water Directive page.