# How a published figure is produced URL: https://municorn.us/methodology/how-we-produce-a-figure/ Updated: 2026-09-08 Tags: methodology, process > **Cite as:** Municorn.us, "How a published figure is produced", Municorn water loss index, https://municorn.us/methodology/how-we-produce-a-figure/ (source filings retrieved 2026-09-08). > **Licence:** CC BY 4.0 — https://creativecommons.org/licenses/by/4.0/ — reuse and redistribution are permitted with attribution to Municorn.us and a link to this URL. > **Caveat that must travel with these figures:** "Water loss" is at least three different quantities depending on the jurisdiction, and nothing here supports a comparison between states. What is kept, what is checked, what is refused and what is never reported. The general statement of the standard every figure on this site is held to; the water loss index is the worked example of it. Four steps, and the fourth is the one that makes the other three worth anything. Everything before it is assembly, which a funded competitor can copy in a few weeks. Declining to state a verdict the record does not support is a posture, and postures do not copy. This is the general form. [The water loss index](/methodology/water-loss-index/) is the same four steps applied to one body of filings, with the specific checks named and the thresholds stated. ## What we keep A copy of each agency file exactly as published, together with the date it was retrieved and a fingerprint of the bytes. Nothing is normalized on the way in and nothing is corrected on the way in. Superseded files are kept, not replaced. Agencies restate, utilities resubmit, and a figure that changed is a fact about the record. See [why historical records matter](/methodology/why-we-keep-versions/). ## What we check **Every published figure resolves to the cell it cites.** A figure on this site carries the file it came from and the address inside that file. A verification pass re-reads the archived bytes and confirms the value published is the value at that address. It runs on every build, and a figure whose coordinate no longer resolves stops the build rather than being published unsourced. **Values that cannot be literally true are nulled and recorded.** A negative volume, a cost per connection that no system could have incurred, a reported loss of exactly zero. These are not corrected to a guess and they are not silently dropped: the field is emptied, the reason is attached to the row, and the page says the value was refused. **Movement is measured against the state's own distribution, not against a number we chose.** For each system we take the mean absolute year-over-year change in its reported figure, and compare it with the same statistic for every system in that state. A series above the state's 75th percentile is reported as unstable — meaning the level cannot be acted on, not that the utility did anything wrong. The cutoff is recomputed from the corpus on every build, so it moves with the state rather than ageing into a constant somebody once picked. **Measured components are separated from defaulted ones, where the regulator publishes the difference.** Some audit forms supply a default value when a utility has not metered a component, and record that they did so. Where that flag is published, we report how much of a figure was measured and how much the form supplied. Texas is the only state in this record whose regulator publishes it; for the others the quantity does not exist, and it is shown as absent rather than estimated. **Cells carrying a note marker instead of a value are carried as the marker.** Some agencies write a symbol into the value cell and explain the symbols in a legend. The marker is preserved and the legend is reproduced, because a marker turned into a number is a number nobody filed. ## What we refuse A filing that cannot be attributed to exactly one system is not published as though it could be. Each refusal is listed, with the reason, on the page for the system it would otherwise have appeared on: - **Two identifiers in one cell.** One form covering two systems, with no way in the file to say which figure belongs to which. - **A malformed identifier**, or none at all. - **Two systems in one county sharing an identifier**, where nothing in the file says which one owns it. A refused filing is listed, never deleted. A row that quietly disappeared is a row nobody can argue with, and the count of refusals is a fact about the agency's file that a program running that form ought to be able to see. ## What we never report **No pass or fail on a filed water balance.** In Texas, water loss is derived from system input minus authorized consumption, so the balance identity holds arithmetically on every filing. A check that cannot fail is not a check, and reporting one would be a verdict with nothing behind it. **No system is "over" a standard this record does not hold.** One regulator in this record has a performance standard we have verified and can cite: MassDEP's 10% unaccounted-for water. Where we hold none, no system is described as failing, exceeding or non-compliant — and that is a limit of this record rather than a finding about the state, which may well have set one we have not read. **No verdict the source documents cannot settle**, and no figure presented as ours when it is the utility's. Where the record is missing, the page says the record is missing. The full list of what this record declines to state, and the rule each refusal follows from, is at [what we do not publish](/methodology/what-we-do-not-publish/). _Municorn.us · https://municorn.us/methodology/how-we-produce-a-figure/ · CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ · source filings retrieved 2026-09-08_ ---