- Article ID: NU54-048
- Section: Hot
- Slug: drinking-water-leakage-measurement
- Prepared: 31 October 2026
- Evidence checked: 31 October 2026
- Dek: A gap between water entering a network and water appearing on bills is not automatically a leaking pipe. Finding the recoverable water requires a defensible boundary, accurate meters and evidence that survives the repair crew's departure.
<!-- ARTICLE_BODY -->
The number that is not a direct measurement
The public picture of water loss is a broken main, a flooded road and a crew digging toward the failure. Much of the water that concerns utilities never produces that scene. Some escapes quietly underground. Some is used legitimately without a bill. Some passes through meters that record too little. Some exists only as a discrepancy between reporting periods or databases. The headline loss estimate mixes these possibilities unless the underlying accounting separates them.
England's latest water-resources assessment makes the stakes immediate. The Environment Agency report updated on October 29, 2026 put reported leakage for 2025–26 at 2,621 million litres a day, broadly unchanged from the preceding year and above the planned level. The report applies to England and cautions that reporting requirements can differ between institutions. It is not a global loss estimate or a reason to treat every utility's percentage as directly comparable. Environment Agency's 2025–26 assessment.
Behind such figures sits an accounting problem with physical consequences. Water entering a defined system must be assigned to legitimate use, storage change, export or loss. But most networks do not have a meter on every possible pathway. Some quantities are measured, some estimated and some calculated as residuals. The apparent precision of the final number can exceed the precision of the inputs.
That does not make water auditing futile. It makes it an engineering discipline rather than a subtraction exercise. A useful audit identifies what is known, what is assumed and which uncertainty is worth reducing first. It then connects the balance to field investigations and verifies that interventions changed the physical system, the revenue record or both. The important question is not only how much water appears to be missing. It is what kind of missing water the evidence supports.
Draw the boundary before subtracting anything
A water balance requires a defined system and period. The boundary might encompass an entire utility, a treatment plant's distribution output or a smaller metered district. Those are not interchangeable. Raw water entering a treatment works cannot simply be compared with customer billing without accounting for treatment processes and the location at which the distribution system begins.
Imports and exports must also be explicit. A utility may purchase water from a neighbor and sell bulk water to another system. If the export is counted as both customer consumption and a subtraction from supply, the balance double-counts it. If an import meter is omitted, the utility can appear unusually efficient because part of its supply is missing from the denominator.
Storage introduces time. A reservoir filling during the audit interval temporarily retains water that has entered but not yet reached users. A reservoir emptying supplies customers with water that entered earlier. Over a long reporting period the net change may be relatively small, but short-interval district analysis can be dominated by it. The physical equation must include the change rather than treating every inflow-outflow difference as leakage.
The same discipline applies to customer boundaries. Leakage before a customer meter and leakage after it can occupy different accounting categories, even though both waste water. The classification depends on the audit convention and responsibility boundary. A utility should state that convention rather than allowing an intuitive phrase such as water lost before use to substitute for a reproducible definition.
A practical boundary document is a map plus a list of meters, tanks, transfers and authorized uses. It identifies who owns each data stream and how missing intervals are handled. Without that document, two analysts can produce different loss figures from the same records while both believe they are following a simple subtraction. The disagreement begins before any calculation: they are balancing different systems.
Real loss, apparent loss and non-revenue water
The standard categories distinguish physical escape from commercial or measurement discrepancies. Real losses concern water escaping from the distribution infrastructure. Apparent losses concern water consumed but not properly measured or accounted for, including relevant meter inaccuracies and data-handling problems. Non-revenue water also includes authorized use that is not billed. These categories help connect each part of the balance to a different remedy. EPA's introduction to water audits and loss control.
A hypothetical annual balance illustrates the distinction. Suppose a defined system receives 10 million cubic metres. Billed authorized consumption is 8 million. Unbilled authorized consumption is estimated at 0.2 million. The remaining 1.8 million is water loss under these assumptions. If apparent losses are estimated at 0.3 million, the residual estimate of real losses is 1.5 million. Non-revenue water is 2 million because it includes the unbilled authorized volume as well as both loss categories.
Those numbers imply different actions. Finding a billing import error can recover revenue without repairing a pipe. Repairing a leak can save physical water without increasing billed consumption. Meter replacement can increase recorded consumption even when customers use exactly the same amount. A utility should not advertise all three changes as identical conservation achievements.
The distinction also prevents unfair assumptions about unauthorized use. A large unexplained balance is not evidence that customers are stealing water. Meter bias, missing accounts, unrecorded flushing and source-meter errors can all contribute. Investigation should follow evidence rather than assign the residual to the most politically convenient category.
The categories are valuable precisely because they resist one headline. A utility can have low physical leakage and poor revenue accounting, or strong billing controls and substantial underground loss. Combining both into a single percentage obscures what management needs to fix. The balance sheet should make those differences visible before investment is chosen.
The largest meter can dominate the entire conclusion
Source and bulk meters deserve disproportionate attention because their errors affect large volumes. A small percentage bias on the system's main inflow can rival the estimated benefit of a major leakage program. Installing thousands of customer sensors does not compensate automatically for an inaccurate production meter at the other end of the equation.
Consider a hypothetical system reporting 20 million cubic metres of annual input. A two-percent input measurement error corresponds to 400,000 cubic metres. If the utility claims a leakage reduction of 200,000 cubic metres while the source-meter correction changed by that much, the physical improvement cannot be inferred from the headline alone. The accounting bridge needs to separate the measurement revision from the operational change.
A calibration certificate is useful evidence, but accuracy belongs to an installed measurement system. Pipe configuration, flow conditions, instrument range, signal processing and maintenance can all matter. The U.S. Bureau of Reclamation's water-measurement guidance emphasizes selection, installation, calibration and continuing verification rather than treating the device specification as the entire accuracy claim. Its manual addresses water measurement broadly, including irrigation; the measurement principle remains relevant without importing every application detail into drinking-water practice. Reclamation's measurement-accuracy guidance.
Testing should cover the operating range. A meter that performs well at a typical daytime flow may behave differently at low night flow or an unusual peak. The audit needs the relationship between error and actual throughput, not merely the best laboratory point. Correction factors should be documented with their evidence and uncertainty.
Meter changes also require continuity checks. Record the old and new readings, installation time, units, scaling factors and any gap. A replacement that improves measurement can create an apparent jump in leakage. That jump may be an honest improvement in knowledge, not a deterioration in pipes. Publishing the explanation protects both the integrity of the trend and the credibility of future savings claims.
Customer meters are a population, not a single accuracy number
Customer-meter error is often represented by one percentage in a top-down audit. That percentage summarizes a population with different ages, technologies, sizes, installation conditions and flow patterns. A defensible estimate needs a sampling strategy capable of representing the volume that population actually measures.
Testing only failed or complained-about meters produces a biased sample. Testing only easy-to-access meters can do the same. A utility might stratify by meter type, age and size, then sample within those groups. Large commercial meters may deserve separate treatment because a small number can account for substantial consumption. The appropriate design depends on the system, but the logic should be documented rather than hidden inside a default factor.
Flow weighting matters. Suppose a hypothetical meter is accurate at high flow but under-registers at very low flow. Its annual error depends on how much customer water passes in each range. Averaging laboratory test errors equally across flow points may not represent annual consumption. Conversely, assigning one household's pattern to all customers can produce false confidence.
A replacement program should distinguish revenue recovery from physical savings. More accurate measurement can increase billed volume without reducing demand. Customers may later change behavior in response to clearer information or bills, but that is a separate effect requiring evidence. The first effect is accounting; the second is consumption.
The utility also needs to verify the path from meter to bill. A precise device can feed a database with the wrong multiplier, unit or account association. A communications failure can cause estimated readings to persist long after actual readings resume. These problems are not solved by buying a more accurate sensor. They require reconciliation between physical registers, telemetry, billing records and customer accounts. An apparent-loss investigation should follow that entire chain rather than stop at the meter's faceplate.
Billing periods can manufacture a seasonal leak
Production data often arrive daily or hourly. Customer bills may be based on staggered monthly, quarterly or less frequent reads. Subtracting one calendar month's billed volume from the same month's production does not necessarily compare water used during the same period. The result can mimic a seasonal leakage pattern even if the pipes have not changed.
A hypothetical example is straightforward. A utility records high summer production, but many customers' summer consumption appears on bills issued in autumn. A simple monthly subtraction shows a summer loss spike and an autumn improvement. Neither is a direct measure of leakage. The analyst must allocate consumption to the relevant service periods or use another defensible reconciliation method.
Estimated bills add another complication. A later actual read can create a catch-up adjustment, making one month appear unusually high or low. Canceled and reissued bills can be counted twice if the extract includes both transactions without their status. Credits are financial events that do not always correspond to negative physical consumption. A water balance should use volume records with clear treatment of adjustments, not blindly sum invoice totals.
The Georgia water-audit manual treats the audit as a structured accounting and data-quality exercise, including the distinction between consumption and loss categories. It is a historical technical manual, not a claim about today's Georgia regulatory deadlines or software version. Georgia's water-system audit manual.
The practical test is to reconcile a sample from original register readings through every transformation to the annual audit input. Check units, read dates, billing status and adjustments. Then examine whether the same logic works for unusual accounts, including bulk customers and properties that changed occupants. A system-wide total can look plausible while concealing offsetting errors. Traceable transactions reveal whether the apparent balance is supported by the records beneath it.
Uncertainty does not disappear in the residual
Real loss is often calculated after other components have been estimated. It therefore inherits their uncertainty. A spreadsheet can display a residual to the nearest cubic metre even when source-meter accuracy, unbilled consumption and apparent-loss estimates are much less precise. Decimal places describe formatting, not knowledge.
For a hypothetical simplified balance, let input be 10 million cubic metres with standard uncertainty of 0.1 million, and authorized consumption 8 million with standard uncertainty of 0.08 million. If the errors are independent, the standard uncertainty of their difference is approximately the square root of 0.1 squared plus 0.08 squared, or 0.128 million. It is not the difference between the two uncertainties. Additional estimated components would add their own contributions.
Independence is an assumption, not a default truth. Two values may share a meter, calibration method or data transformation. Their errors can move together. Systematic bias can persist through repeated years, so averaging many observations does not necessarily remove it. A rigorous assessment distinguishes random variability, uncertain correction factors and known data defects.
The purpose is not to paralyze action until every input is perfect. It is to identify which decisions the evidence can support. If a suspected loss is far larger than plausible measurement error, field investigation may be clearly justified. If two investment options differ by less than the uncertainty in their estimated savings, the ranking deserves caution or additional measurement.
Report uncertainty in a form decision-makers can use. Explain which inputs dominate it and what a targeted test could resolve. A source-meter verification might be more valuable than another month of elaborate modeling. A good audit turns uncertainty into an investigation plan. A poor audit hides it behind a single confidence-sounding score and then treats every movement in the residual as a physical change.
District metering changes the scale of the question
A whole-system balance says how much water may be missing across the utility. It rarely says where to dig. District metering divides the network into smaller areas with known inflows and outflows, allowing operators to compare local patterns and prioritize investigation. The advantage comes from a more informative boundary, not from the word smart attached to a meter.
The boundary must actually hold. An undocumented open connection between districts can make one appear to lose water while another appears unusually efficient. Valve status, bypasses, temporary supplies and construction changes need to be reflected in the analysis. A district map that is correct on paper but wrong underground undermines every subsequent calculation.
Storage inside the district must be included. A tank filling at night increases inflow without indicating a leak. A tank supplying the district while an inlet meter records little flow can create the opposite illusion. Synchronized tank-level and flow data allow the balance to distinguish these states. Missing or poorly calibrated level measurements can limit the result even when the inlet meter is excellent.
EPA's technical manual describes district metered areas and night-flow analysis as bottom-up tools that complement the annual audit. The essential idea is to compare measured flow with legitimate use in a defined area, rather than infer the location of losses from the total system residual alone. EPA's distribution-system loss-control manual.
District design also has operational constraints. Closing valves can affect pressure, resilience, fire-flow capability and water movement. A leakage investigation should not casually create a less reliable supply network. Qualified utility engineers need to evaluate those effects. The measurement boundary is valuable only if it can be maintained without compromising the service the network exists to provide.
Minimum night flow is not all leakage
Night-flow analysis works because many legitimate demands are lower during quiet hours. Lower demand can make a persistent leak easier to distinguish. But the minimum measured flow still includes real use: industrial shifts, hospitals, cleaning, irrigation, household activity and any other demand occurring in the selected interval. Treating the entire minimum as leakage overstates the answer.
A hypothetical district records a stable minimum of 12 litres per second. Estimated legitimate night use is 4 litres per second, leaving an 8-litre-per-second residual before other corrections. If the legitimate-use estimate is actually 6, the inferred leakage falls to 6. The measurement may be precise while the interpretation remains uncertain. That uncertainty should guide customer sampling or operational checks.
The selected night matters. A holiday, unusual industrial schedule or irrigation event may not represent ordinary conditions. Rather than choose the lowest number that supports a preferred story, examine a time series and document exclusions. Persistent patterns are more informative than an isolated minimum. A sudden step change can indicate a leak, but it can also indicate a valve operation or meter problem.
Converting night leakage into a daily volume requires another assumption: how leakage changes with pressure across the day. Multiplying the night residual by 24 hours can be misleading if pressure differs substantially. The calculation should use a justified pressure relationship or state the simplification and its likely effect.
Night flow is therefore a diagnostic instrument, not a verdict. It can narrow the search and provide a baseline against which repair effects are assessed. It cannot, by itself, tell the operator whether the residual comes from one large leak, many small leaks or an unrecognized legitimate demand. Field evidence has to complete the interpretation. The more clearly the assumptions are recorded, the easier it is to update the estimate when new evidence arrives.
Pressure management changes both losses and the baseline
Pressure affects leakage, but the relationship depends on the defect and the network. A rigid opening and a crack that expands under pressure do not necessarily respond in the same way. Pressure management can be valuable, but a universal percentage saving should not be assumed from a percentage pressure reduction.
The Texas Water Development Board's technical manual describes pressure management as a way to address excessive leakage and pressure-related stress, while noting that not every utility has the same opportunity. Its discussion supports investigating the operating range and transients rather than treating lower pressure as an unconditional goal. Texas water-loss audit manual.
A simplified hypothetical relationship is leakage proportional to pressure raised to an exponent. If the exponent were one, a ten-percent pressure reduction would imply a ten-percent leakage reduction under otherwise unchanged conditions. If the exponent differed, so would the response. This is an illustrative sensitivity calculation, not a site-specific design rule. The relevant relationship needs evidence for the system and operating range.
Service requirements remain constraints. Pressure at a convenient monitoring point does not necessarily represent the highest elevation, the end of a long branch or a peak-demand condition. A program must preserve adequate service and applicable safety requirements. Operators should also consider transients rather than only average pressure; abrupt changes can matter even when the daily mean looks acceptable.
Pressure changes complicate before-and-after verification. A repaired leak may restore pressure elsewhere, increasing flow through remaining defects. Conversely, lower pressure may reduce total leakage without any pipe repair. Neither outcome invalidates the intervention, but the accounting should distinguish the mechanism. Otherwise a contractor can claim all observed reduction as repair success, or a useful repair can appear disappointing because the rest of the system responded physically to the change.
Finding a leak and measuring its savings are different tasks
Acoustic sensors, correlators, inspections and other detection methods can locate defects that are difficult to see. Their performance depends on the network and conditions. A tool's ability to flag a likely leak does not automatically establish the leak's volume, duration or recoverable annual saving. Those are additional questions.
A confirmed location should feed a structured work record. Record discovery time, evidence, pipe or connection details, estimated flow, pressure conditions, repair time and post-repair observations. Distinguish the date the leak was reported from the date it probably began. A visible burst may be repaired quickly but lose less total water than a small hidden leak that persists for months.
A hypothetical comparison makes this intuitive. A leak flowing at 10 litres per minute for a year loses about 5.26 million litres. A burst flowing at 1,000 litres per minute for two hours loses 120,000 litres. These simplified calculations assume constant rates and are not estimates of a real network. They show why duration belongs beside flow in the priority discussion.
Detection programs should report false alarms and unresolved signals as well as confirmed repairs. A high number of alerts can reflect sensitive monitoring, noisy data or poor targeting. The operational value depends on how many actionable defects are found, how quickly they are addressed and what measurable benefit follows. A map full of colored pins is not a water-saving result.
The same caution applies to machine-learning detection. A model trained on known bursts may not perform equally well on slow background leakage. Evaluation should distinguish event types, pipe environments and missing sensor conditions. Independent field confirmation and measured post-intervention changes are stronger evidence than a model's confidence score. The technology helps direct attention; the repair and the water balance establish what changed.
The percentage can improve while leakage gets worse
Loss percentages depend on their denominator. If customer demand rises while physical leakage remains constant, leakage as a share of input can fall. If customers conserve water while leakage remains constant, the percentage can rise. A single ratio can therefore reward increased consumption or make conservation appear to worsen network performance.
Imagine a hypothetical utility with 2 million cubic metres of real loss and 8 million of legitimate use. Real loss is 20 percent of the 10-million total. If legitimate use falls to 6 million while loss remains 2 million, the share becomes 25 percent. The network did not begin leaking more. The denominator changed. An honest performance report presents absolute volume alongside the percentage.
Other normalizations answer other questions. Loss per service connection can help compare systems with different customer counts. Loss per length of main may be more informative where connections are sparse. Pressure-adjusted indicators can provide additional context, but they depend on the quality of the pressure and infrastructure data used. No indicator makes unlike networks automatically comparable.
Comparisons should state the reporting period and methodology. A revised estimate of customer-meter error can shift calculated real loss without changing a single pipe. A boundary expansion can add network length and customers. A move from estimated to measured consumption can alter the baseline. Trend charts need a bridge explaining these changes rather than presenting a continuous line as though its meaning never changed.
This is not an argument against targets. It is an argument for targets that resist easy denominator tricks and remain interpretable when demand changes. Decision-makers need a small set of complementary measures: physical volume, financial effect, service performance and data quality. Together they show whether the utility is improving the network, improving its knowledge or merely changing the arithmetic used to describe it.
Put the right value on the recovered water
Physical water saved and revenue recovered have different economic values. A real-loss reduction may avoid pumping, treatment or purchased-water costs. It may also defer capacity investment or reduce pressure on a constrained source. An apparent-loss correction may recover billable revenue from water already consumed. Applying the retail tariff indiscriminately to every litre can overstate some benefits and understate others.
A hypothetical leak repair saves 100,000 cubic metres a year. If the short-run avoidable production cost is 0.40 currency units per cubic metre, the immediate operating saving is 40,000 units under that assumption. Multiplying by a retail tariff of 2.00 would produce 200,000, but that is not automatically cash the utility gains. The saved water may not be sold, and parts of the tariff cover fixed costs that remain.
The longer-term value can nevertheless exceed short-run operating savings. A constrained system may use saved water to delay an expensive supply project or improve drought resilience. Those benefits require a planning model and a credible counterfactual. They should be stated separately rather than hidden inside an inflated unit price.
Apparent-loss recovery also needs realism. Correcting a meter can increase recorded consumption, but collection rates, tariffs and customer protections affect the realized financial result. A billing-system correction should be checked for fairness and accuracy, not pursued as though every increase in billed volume proves previously uncharged consumption.
Investment comparisons should use consistent time horizons, discounting assumptions and maintenance costs. A cheap repair with a short life and a more durable renewal are not equivalent simply because their first-year savings match. Include investigation costs, traffic management, disruption and the possibility that leakage will reappear elsewhere. The balance sheet becomes useful when it connects physical evidence to a defensible economic decision, not when it turns every uncertain litre into a large headline valuation.
Verify savings against a counterfactual, not a photograph
A photograph of a repaired pipe proves that work occurred. It does not prove the annual savings claimed for the program. Verification needs a baseline, an intervention record and a method for estimating what would have happened without the intervention. Weather, demand, pressure and measurement changes can all influence the observed balance.
For a small district, a clear step reduction in comparable night flow after a confirmed repair can be persuasive evidence. The comparison should account for legitimate night use, tank operations and pressure. For a system-wide program, the signal may be distributed across many interventions and obscured by new leaks. A measured flat trend can represent substantial avoided deterioration, but that claim requires an explicit and testable baseline model.
Avoid double-counting. If a leak repair and a pressure-management project affect the same area, their savings are not necessarily additive. Each may change the baseline available to the other. A contractor paid for both should not receive credit twice for the same reduced flow. The measurement plan should define attribution before work begins.
Persistence matters. A reduction measured immediately after repair may not last through a full operating cycle. Follow-up should check whether the improvement remains, whether new defects emerge and whether the instrumentation continues to work. A program's value is the durable reduction in loss, not the best single day on the chart.
California's current water-loss program provides a concrete example of recording more than a headline. Its official page lists individual system standards released in January 2026 and a leak registry covering breaks, repairs and estimated losses, with the first registry covering 2025–27 and due in 2029. These are California-specific requirements and dates. The broader analytical lesson is the value of connecting annual balances to event-level evidence. California's water-loss-control program.
A useful audit tells the utility what to measure next
The first audit of a poorly documented network may have wide uncertainty. That is not a reason to postpone it indefinitely. A structured first pass can reveal which unknowns dominate the result and which investigation would most improve the next decision. The audit is valuable as a map of ignorance as well as an estimate of loss.
Suppose the uncertainty is dominated by one unverified bulk meter. Testing that meter may be a better first investment than deploying many leak loggers. Suppose billing records are sound but a district shows persistent unexplained night flow. Field investigation becomes more compelling. Suppose a large share is unbilled authorized use with weak records. Metering or logging those uses can clarify the balance before physical-loss estimates are refined.
AWWA's free audit software supplies a standardized top-down framework. Its current page identifies version 6.1 as adding carbon calculations to the earlier functionality. The tool helps organize the balance; downloading a workbook does not constitute a comprehensive field investigation or validate every input automatically. AWWA's free water-audit software.
The utility should maintain an improvement register tied to specific data fields. Each item needs a question, a measurement method, an owner and a decision it will inform. Replace vague instructions to improve data quality with tasks such as verifying a meter under representative flow conditions or reconciling a sample of billed volumes to original reads.
Progress can then be reported in two dimensions: reduced loss and improved confidence. A revised audit may initially show more leakage because previous assumptions understated it. That is uncomfortable but useful knowledge. Penalizing every upward revision encourages institutions to preserve optimistic estimates. A mature program rewards accurate accounting while still demanding physical improvement. The balance has to be allowed to tell an inconvenient truth before it can guide a successful repair.
Carbon accounting needs the same physical boundary
Water-loss programs increasingly discuss energy and emissions alongside volume. That can be useful, but the carbon calculation inherits the same boundary and counterfactual problems as the water calculation. A litre saved in a gravity-fed section does not necessarily avoid the same energy as a litre that would have been pumped through several stages and treated intensively.
The calculation should identify which processes actually change when loss falls. Some treatment and pumping energy varies with throughput; some remains because equipment must continue operating. The electricity emissions factor can vary by place and time. Purchased water may carry upstream energy that is not visible in the utility's own power bill. A broad average can be suitable for screening, but it should not be mistaken for a precise project result.
A hypothetical project that saves 50,000 cubic metres and avoids 0.5 kilowatt-hours per cubic metre would avoid 25,000 kilowatt-hours under that assumption. Converting this to emissions requires an explicit factor and an explanation of whether it represents average or marginal electricity. The arithmetic is straightforward; establishing the avoided energy is the harder evidentiary step.
Do not count revenue recovery as physical water savings. A more accurate meter may change the balance category without reducing treatment or pumping. If customers subsequently reduce consumption, that effect can be measured separately. Combining both into one carbon claim risks crediting emissions reductions that never occurred.
Emissions can strengthen the case for a good leakage project, but they should not rescue a weak water balance through multiplication. The most credible report keeps the chain visible: verified physical reduction, affected processes, avoided energy and chosen emissions assumptions. Readers can then see which parts are measured and which are modeled. The carbon result becomes an extension of the evidence rather than a second layer of impressive precision placed on top of an uncertain first estimate.
A final reconciliation should check units explicitly. Litres per second, cubic metres per day and annual megalitres can all describe the same flow, but an unnoticed conversion error can overwhelm a genuine improvement. Preserve original units in the source record, perform conversions in a documented step and test the results against a simple independent calculation. The same caution applies to calendar length: a leap year, an incomplete reporting interval or a meter outage changes the volume represented by an average daily rate.
Responsibility for that reconciliation should sit with someone able to question both operations and finance. Neither team necessarily holds the complete picture. Operators know why a tank was drained; billing staff know why a large account was adjusted. Bringing those explanations into the same record prevents an otherwise careful audit from treating ordinary, documented events as mysterious losses. The final number should be the product of reconciliation, not merely the output of a spreadsheet formula.
The balance sheet should survive the press release
A defensible water-loss claim can be followed backward. Start with the reported improvement, identify the system boundary and period, inspect the measured inputs and estimated components, examine the corrections and trace the interventions that plausibly caused the change. The numbers need not be perfect. They need to be coherent, reproducible and honest about their limits.
That standard changes the public conversation. A utility can explain that one project repaired physical leaks, another improved meter accuracy and a third reduced uncertainty about an important source. It can show which benefits are water savings, which are revenue recovery and which are better knowledge. None needs to be disguised as the others to be worthwhile.
It also changes procurement. Detection suppliers should be evaluated on confirmed useful findings and operational outcomes, not just alerts. Metering projects should demonstrate measurement and data-chain performance. Repair contracts should define how savings will be verified and how overlapping interventions will be handled. The audit becomes the common language connecting these purchases.
The most important figure may therefore be less dramatic than the headline percentage. It may be the uncertainty on a source meter, the legitimate-use estimate behind a night-flow calculation or the number of days between detecting and repairing a persistent leak. Those details determine whether a large claim reflects water retained in the system or a change in how the spreadsheet classifies it.
Water conservation deserves reliable accounting because the physical resource is real even when the estimate is indirect. A missing litre cannot be repaired as an abstract residual. It has to be assigned a plausible pathway, investigated with the right tool and checked after the intervention. The balance sheet is where that process begins, and where the claimed success should eventually return. <!-- /ARTICLE_BODY -->