The Infrastructure Leakage Index, explained.

Ask two water utilities how much they leak and you will get two percentages that cannot be compared. A dense city network losing 15% of system input and a rural network losing 15% are not in the same condition — the rural one, with kilometres of mains per connection, is doing considerably better. Percentages describe how much water moved, not how well a network is being run.

The Infrastructure Leakage Index exists to fix that, and it is the method the EU’s Drinking Water Directive names for assessing leakage under Article 4(3). If your national authority is asking for a leakage figure, this is the shape it will eventually take.

What the index actually is

ILI is a ratio of two numbers:

ILI = CARL / UARL

• CARL — Current Annual Real Losses. The water your network actually loses through holes: burst mains, weeping joints, leaking service pipes. Real losses only; meter error and unbilled-but-delivered water are a different category.
• UARL — Unavoidable Annual Real Losses. A theoretical floor: what a network of that exact shape would still lose if it were in excellent condition and every current best practice in leakage control were applied to it.

So an ILI of 1.0 says a network is losing roughly as little as its physical layout allows. An ILI of 4.0 says it is losing four times what a well-run network of the same geometry would.

The second number is what makes the index work. Because UARL is computed from the network’s own mains length, connection count and pressure, it adapts to the network being measured — which is exactly what a percentage of system input refuses to do.

How UARL is calculated

The formula was proposed by Allan Lambert in 1999 and adopted by the International Water Association:

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

Where:

• Lm — length of mains, in km
• Nc — number of service connections
• Lp — total length of underground service pipe from the property boundary to the meter, in km
• P — average operating pressure, in metres of head

The three coefficients are leakage allowances for three different places water escapes: the mains themselves, the connections where they join, and the service pipes running to the meter. Everything is multiplied by pressure, because leakage scales with it — the same hole passes more water at 60 m of head than at 30 m.

That pressure term is the one people get wrong. It is average operating pressure across the zone, over the period, weighted properly. Assume it rather than measure it and you move the ILI without anything changing in the ground.

A worked example

A mid-sized municipal network:

UARL = (18 × 120 + 0.8 × 4,000 + 25 × 8) × 45 UARL = (2,160 + 3,200 + 200) × 45 UARL = 5,560 × 45 = 250,200 litres/day, or about 250 m³/day

Now suppose the water balance puts current real losses at 700 m³/day — roughly 255,500 m³ a year.

ILI = 700 / 250 = 2.8

That network is losing about 2.8 times what its geometry makes unavoidable. Note what the number does not depend on: how much water the town consumes. A dry summer that pushes consumption up does not flatter the ILI, and that stability is the entire point.

What counts as a good ILI

Lower is better, and 1.0 is the practical floor rather than a target — getting there costs far more than the water is worth for most utilities.

Beyond that, be careful with league tables. Published ILI bands exist and get quoted confidently, but what is economically sensible depends on the price of water, the cost of finding leaks in that terrain, energy prices and how stressed the source is. A network in a water-scarce region should be chasing a lower ILI than one beside a large river, and both can be making the right call.

The more useful comparison is with yourself. An ILI computed the same way each year, trending down, says more than a single number benchmarked against someone else’s network.

Where the index stops being trustworthy

ILI is a good metric inside its range and a misleading one outside it. Four cases to watch:

• Small systems. The UARL formula was derived from data on large networks. On small ones — few connections, short mains — it produces a denominator too small to be reliable, and the resulting ILI swings wildly on small changes. Lambert’s own guidance flags this. If your network is small, compute litres per connection per day as well and trust that more.
• Low pressure. The formula assumes leakage scales with pressure across a normal operating range. At unusually low average pressures the linearity breaks down and UARL is understated.
• Real losses contaminated by apparent ones. This is the common failure. If your CARL still contains meter under-registration, unauthorised consumption or data-handling errors, you are dividing the wrong numerator by a correct denominator and calling the result an ILI. An ageing meter fleet that under-reads at low flow will inflate your apparent leakage substantially.
• Guessed inputs. Mains length from a GIS layer nobody has reconciled, connection counts from a billing system that still lists demolished properties, pressure taken from a design document. Each of these moves the index and none of them move the water.

What you need to compute one honestly

Four things, and the first three are unglamorous:

1. A water balance you believe. System input measured at every source, billed consumption from a meter fleet you know the condition of, and unbilled-but-authorised water actually written down rather than estimated.
2. Real and apparent losses separated. Otherwise the numerator is wrong.
3. Asset data that matches the ground. Mains length, connection count and service pipe length, reconciled at least once before the first calculation.
4. Measured pressure. Zone-average, over the period, not a nominal figure from a drawing.

Water loss management on Divako keeps the first two continuous rather than annual — a balance per zone, recomputed as readings land, with real and apparent losses held apart. The asset data and the pressure are yours; what the platform can do is stop the numerator being guesswork.

If you are working toward a leakage assessment, the Drinking Water Directive page covers the obligation, who is in scope and the 2028 timetable.