Methodology

How a number gets to the screen

Four steps, and only the second involves any judgement at all.

1. 1You enter a measurement. Two meter readings and two times, a drip count, a pool level drop, a figure off a bill.
2. 2The calculator applies constants. Every constant is declared once, with a value, a unit, a source and a link to that source, and no number is written directly into a calculation.
3. 3The arithmetic runs. Unit conversions are definitional and identical for everyone: 1,000 litres to a cubic metre, 24 hours to a day, 365 days to a year.
4. 4The result renders the same six figures, in the same order, on every page: rate of loss, daily volume, annual volume, cost, severity, next action.

The consequence of step two is that there is no such thing as a figure on this site whose origin cannot be found. If a number appears in a result, it came from a constant with a source, from something you typed, or from multiplying those together.

One thing to know before you check any of it by hand. The calculators work at full precision and round only when they display, and how many decimal places each kind of figure is shown to is a single published contract rather than a choice each page makes. That is so two pages showing the same quantity cannot show it to different precision.

Worked examples in the prose show the exact arithmetic instead. So if you multiply a rate out yourself you may land on a final digit the tool does not display, and that is rounding rather than disagreement. The rounding is deliberately coarser than the arithmetic in places, because a tenth of a litre a day is well below anything a domestic water meter can resolve, and displaying a digit the instrument cannot support would be inventing precision.

There is one consequence of that worth pointing out before you find it. The cost is calculated from the unrounded annual volume, not from the rounded one shown on screen. So if you take the annual volume the page displays and multiply it by your own rate, you can land a few pence away from the cost the page displays.

Both of those figures are right. It is the arithmetic between two already-rounded numbers that is wrong, which is a general property of rounding rather than anything specific to this site. Rounding once at the end is more accurate than rounding twice on the way.

Where a figure has no source

Some figures cannot be sourced, because nobody has published a measurement of them. Those are labelled assumptions and handled differently from constants.

• An assumption is never hidden inside a calculation. It renders as an input field with its default filled in.
• The default is stated, along with what it is: a convention, a trade rule of thumb, or a judgement.
• Changing it changes the result immediately, so the sensitivity of the answer to that assumption is visible rather than described.

There are 11 of them by the count on the sources page, between a quarter and a third of the graded figures there, and they are listed with everything else on every citation on this site, graded and ordered worst first.

Letting you override an assumption is more honest and more useful than pretending the assumption is a fact. It is also the only way a stranger can tell how much of an answer is measurement and how much is us.

Where the physics does not support a precise answer

Three places on this site deliberately refuse to give a number that a competing tool will give you, and each refusal is the more useful output.

What we found when we tried to trace the figures everyone else uses

Two of the most widely repeated water-saving figures in the UK are published without a stated method. Both come from bodies that do genuinely useful work, and this section reports what their pages contain and do not contain rather than making a claim about whether the numbers are right.

The figure is reachable. At 5,500 litres a year the rate is about 15 litres a day, which needs roughly 42 drips a minute at the only drop volume anyone has published from measuring real taps. That is a fast, badly worn tap, and it is entirely plausible. So "up to" is a fair reading of the claim rather than a rebuttal of it. What is absent is the working, not the plausibility.

A measured alternative does exist for the second of those. A survey attributed to Thames Water, reported by a third party rather than published directly, inspected 58,551 toilets and found 4,854 leaking, which is 8.3 per cent. That figure has a real sample behind it and it is what this site uses, cited through the intermediary that was actually read rather than as though the original had been.

The two most-repeated UK household water loss figures are both published without a stated method, and the same organisation is a party to both publications. That is not two independent confirmations, and this page will not present it as one. It is a fair observation about one body's publishing practice, on the two figures most likely to be quoted back at you, and it is worth knowing on exactly that basis and no wider.

It is also why this site publishes its constants. A figure whose method is published can be argued with. A figure whose method is not published can only be repeated, which is precisely what has happened to both of these.

Machine-readable output

Each page carries structured data describing the page, the calculator where there is one, the breadcrumb trail, the questions and answers, and the measurement procedure where a genuine one exists.

The questions and answers in that markup are not a second copy of the visible text. They are the same object, rendered twice, and a build check compares them character for character and fails if they differ. That matters because a machine extracting an answer should get the answer a reader gets.

This is done so a sourced number can be extracted correctly by something reading the page. No claim is made about search result features, and none should be inferred.

What the build checks do and do not prove

This site is built with a set of automated checks that refuse to publish it if certain things are wrong. They enforce the rules on this page: that the markup matches the visible text, that no page invents its own rounding, that a figure nobody could source is exposed as an adjustable input rather than hidden as a constant, and that several things the site has promised never to do are not done.

They are worth being precise about, because it would be easy to imply more than they deliver.

Every one of those checks is structural. They can tell that two copies of a sentence match, that a date is present, that a number is displayed to the agreed precision. **None of them can tell whether a number is true.** A wrong figure, confidently transcribed and carrying today's date, passes every check on this site.

So a green build says the site is internally consistent. It does not say the site is right. The only thing that speaks to whether a figure is right is where it came from, which is why every constant carries its source and why the sources page grades the strength of each one rather than simply listing it.

Assumptions and sources

Every constant this site uses is listed above with its value, its unit and where it came from. This page rests on the whole of that rather than on any particular figure, so the full list is shown rather than a filtered view that would look complete and would not be.

Common questions