Water utilities: separate leakage and unbilled consumption
A water utility can improve billed consumption without saving any physical water. It can also reduce leakage without increasing bills. These are different operating mechanisms with different evidence and economic values. Research should first close the volume balance, then identify which category changed before interpreting a non-revenue-water percentage.
The deliverable is a four-category water ledger and two separate intervention cases. It covers water entering a defined distribution system during a defined period, with storage changes reconciled. It does not estimate river-basin consumption or certify an individual utility’s engineering performance. All numbers below are fictional, and dollar values depend on expressly assumed prices and collection.
Locate the category definitions
EPA’s Water Audits and Water Loss Control for Public Water Systems introduces authorized consumption on printed page 1 and distinguishes real losses, apparent losses and non-revenue water on page 2. Real losses concern physical leakage; apparent losses concern consumption missed through unauthorized use, metering inaccuracies or data-handling errors. Unbilled authorized consumption is another component of non-revenue water.
Use those categories to structure the worksheet. The document’s Water Audit, Intervention and Evaluation framework supports checking results after a change. It does not supply current company loss rates, recovery guarantees or the invented economics below. Older introductory estimates in the document should not be presented as a current national investment requirement.
Ask for the system-input meter, billed consumption ledger, authorized unbilled-use records, customer-meter tests and loss-estimation method for the same period. Record purchased bulk water and transfers out so the distribution boundary is consistent. Do not subtract a tank’s closing stock from an annual flow without also incorporating its opening stock.
Reproduce a complete baseline
Assume annual system input of one hundred million cubic metres. Billed authorized consumption is seventy, unbilled authorized consumption five, apparent losses ten and real losses fifteen, all in millions of cubic metres. No inventory change or other stream is omitted in this teaching case.
The volume identity is 100 = 70 + 5 + 10 + 15. Non-revenue water is 5 + 10 + 15 = 30, or 30% of input. Water losses alone are apparent plus real, equalling twenty-five. Calling all thirty leakage would wrongly include authorized unbilled use and apparent losses.
| Category | Million cubic metres/year | Evidence object |
|---|---|---|
| Billed authorized use | 70 | Customer billing volumes |
| Unbilled authorized use | 5 | Documented authorized activities |
| Apparent loss | 10 | Meter, data or unauthorized-use evidence |
| Real loss | 15 | Physical-loss estimate and supporting measurements |
| Total system input | 100 | Matched input measurement |
If apparent loss is estimated, real loss may be the remaining balance rather than an independently measured number. Label it accordingly. A closed arithmetic identity does not prove every category’s accuracy. Preserve meter uncertainty and the method used to allocate unmeasured quantities instead of treating the residual as proven theft or pipe leakage.
Value a physical repair case
Assume a verified repair reduces real leakage by five million cubic metres, and the utility reduces input by the same amount while serving unchanged demand. Input falls to ninety-five; billed use remains seventy, authorized unbilled use five, apparent loss ten and real loss ten. The new balance closes at ninety-five.
Non-revenue water becomes twenty-five, or approximately 26.32% of input. Under the explicit assumption that each avoided cubic metre saves $0.40 in marginal production and purchase costs, annual gross avoided cost is $2 million. This is before repair expenditure, maintenance, financing and any other incremental cost.
Do not multiply the saved water by the retail tariff and also credit the full avoided production cost unless the scenario establishes both separate benefits. In this case demand and billed volume are unchanged, so the direct modelled benefit is avoided input cost. If freed water can instead serve additional paying customers, build a different volume and incremental-cost case using that evidence.
Value a meter-correction case
Return to the original baseline and assume meter correction reclassifies five million cubic metres from apparent loss to billed authorized consumption. Input remains one hundred, billed use rises to seventy-five, authorized unbilled use remains five, apparent loss falls to five and real leakage remains fifteen. Non-revenue water is twenty-five, or 25% of input.
Physical abstraction and physical leakage have not decreased in this example. The change is better measurement and billing of water already consumed. With an assumed incremental tariff of $2 per corrected cubic metre and 90% collection, incremental collected revenue is 5 million × $2 × 90% = $9 million, before intervention and collection costs.
The tariff and collection factor are assumptions. A meter-test discrepancy does not automatically become a collectible customer balance. Contract terms, billing rules, timing and actual receipts need separate evidence. Keep newly measured volume, invoiced value and cash collected in separate columns; do not convert every apparent-loss correction directly into cash.
Test the percentage denominator
Now consider a different boundary case: billed demand falls from seventy to fifty while all other categories remain unchanged. Input becomes eighty and non-revenue water remains thirty. The non-revenue percentage rises from 30% to 37.5%, although physical leakage remains fifteen. A worse percentage alone does not prove more cubic metres leaked.
The reverse can also occur if expanding billed demand lowers the percentage without repairing any pipe. Report absolute volume, the input denominator and the physical-loss estimate together. Compare similar system boundaries and observation periods before assigning improvement to management action.
If both original interventions occur independently with the stated volume effects, input is ninety-five, billed use seventy-five, unbilled use five, apparent loss five and real loss ten. Non-revenue water is twenty, about 21.05%. This combined case is a new ledger, not the sum of two percentage-point improvements calculated on different denominators.
Apply the verification sheet
| Input | Obtain | Check |
|---|---|---|
| System input | Calibrated meters and boundary map | Purchases, transfers and storage movements |
| Billed volume | Consumption ledger and meter dates | Same period, no cash/volume substitution |
| Unbilled use | Authorized records or labelled estimates | Legitimate unbilled activity not called leakage |
| Apparent loss | Meter tests and billing reconciliation | Sampling basis and extrapolation stated |
| Real loss | Estimation method and physical observations | Residual distinguished from direct observation |
| Intervention value | Marginal costs, billing and receipts | Avoided cost distinct from collectible revenue |
When two reports disagree, first reconcile the system boundary, meter dates, purchased water, tank movements and category revisions. Do not average a financial-year balance with a calendar-year balance or assign the whole difference to unauthorized use. Retain a named unresolved residual and specify the record needed to explain it.
Audit evidence quality before comparing tiny improvements. If input meters are replaced, establish whether the apparent change comes from corrected measurement rather than physical savings. A before-and-after ledger should preserve the original measurements and a comparable restated series, with the restatement reason recorded.
Update when meters, customer counts, authorized uses, tariffs, collection or system boundaries change. The conclusion should distinguish water physically saved, consumption newly measured and bills actually collected. A lower non-revenue percentage is useful only after those objects and the denominator can be reproduced.
Separate measurement restatement from an intervention
Suppose a later input-meter check restates the original input from one hundred to ninety-eight million cubic metres, while billed use, authorized unbilled use and apparent-loss estimates remain seventy, five and ten. A residual-based real-loss estimate becomes thirteen rather than fifteen. That change alone does not establish two million cubic metres physically saved: the earlier input measurement changed.
Retain three versions where possible: originally published balance, comparable restated baseline and measured post-intervention balance. Attribute the baseline revision separately from the repair result. If only the revised annual total is published, the two effects cannot be distinguished from that number alone.
For customer-meter studies, obtain the population tested, selection method, meter ages, consumption weights and treatment of failed readings. An average error across meters is not necessarily the volume-weighted error across water consumed. A small customer and a large user should not receive equal volume weight merely because each owns one meter. Without a defensible extrapolation, report the observed test result and leave system-wide apparent loss as an uncertain estimate. This prevents a selective meter sample from becoming a precise revenue-recovery claim.
FAQ
Is all non-revenue water physical leakage?
No. It also includes apparent losses and authorized unbilled use in this framework.
Does replacing meters prove lower water withdrawals?
No. It may improve billing while the physical water balance remains unchanged.