About this site

One engine, not many pages

Every calculator is a declaration: its inputs, a pure function that turns those inputs into a result, the standards it cites, its test cases and its FAQ. A single renderer turns any of those declarations into a working page.

The practical consequence is that an improvement lands everywhere at once. When the unit handling was corrected, every calculator was corrected. There is no page carrying an old version of a shared behaviour, because there is only one copy of it.

Every formula names its source

A result you cannot check is a result you should not use. Each calculator lists the standards behind it with the clause or section number, so the arithmetic can be traced back to a document that has authority — ACI, IRC, NEC, IPC, ASHRAE, Eurocode, AS/NZS and BS, depending on the subject and the market.

Where no standard exists, or where the honest answer depends on data the site does not have, the page says so instead of estimating. That is why some calculators state a limitation rather than producing a confident figure: a plausible number in that position is worse than an absent one.

What lastVerified means

It is the day the page was written, or last revised, with its cited sources checked against the calculation at the same time — not the date it was published as such, and not a promise that the standard has not changed since. Pages written in the same wave and not changed since carry the same date, and that is all the sameness means: they were authored that day, not revisited one by one later. A page corrected in the 2026-09 audit carries the date of that correction, and its changelog, printed on the downloadable report, says what changed.

The second date, labelled 'in the site-wide review', is a later visit. On that day every calculator then on the site was read again, individually, in a first-party review — by this project's own reviewers, not an independent body: its arithmetic, units, bounds and stated sources were checked, each wrong-number finding was re-checked by two further reviewers working separately, and the confirmed ones were corrected. It says the page was read in that review, not that its working is unchanged since: a page revised afterwards may differ from what was read. It appears only on a page whose sources were last checked on or before the review; a calculator written after it was not part of it and shows its own date alone.

Codes move on cycles. A clause that was current when checked can be superseded, and nothing here notices automatically. Treat both dates as the point from which you verify anything load-bearing yourself.

Pinned test cases

Each calculator carries worked examples with expected results, frozen as automated tests. Any change that moves one of those results fails the build.

The point is not to prove the formula is right — a test can only prove it still does what it did. It is to make silent drift impossible: a refactor, a unit change or an overzealous tidy-up cannot quietly alter a published figure.

Confidence, stated on the result

Results carry a confidence level. Arithmetic on measured dimensions is high. An estimate from an empirical relationship with real scatter is medium. A first-order approximation that ignores factors which often dominate in practice is low — and says what it ignores.

A mass-law sound-transmission estimate is a good example: it models a single solid leaf and knows nothing about decoupling, cavity depth or resilient channels, which in a real partition frequently matter more than mass. It is useful for comparing options and useless for certifying a party wall, and the page says exactly that.

A machine check against thin, templated content

Generating many pages from one engine carries an obvious risk: pages that read as variations of each other. Every build measures the similarity of every page against every other page, discounting shared site furniture, and fails if any pair crosses a threshold.

That check has changed the writing more than once. An earlier version of the derived commentary used one fixed sentence structure, and the audit caught hundreds of page pairs that were too alike. The fix was to write better, not to raise the threshold.

Every link runs both ways

A calculator is rarely the whole question, so its page names the pages that take the question further: the guides that use it, the comparisons it is one side of, the glossary terms for what it measures, the paper that explains its method, the standards it cites, the job takeoffs and the planner it is part of, the diagnoses that point to it, the worked examples that run it, the reference tables its figures come from and the field tools that can take its measurements. Each of those pages names it back.

Every build reads each page it publishes, collects both ends of every one of those links, and stops if a link is on one page and not the other, or if the table below disagrees with what the pages carry. The figures are counted from the built pages each time the site is published; none of them is typed.

One kind of page is left out on purpose. A calculator links to the entry for each tool a job needs in the equipment catalogue, the way a dictionary is used, and the catalogue names no calculator in return, because a tape measure serves nearly all of them.

What this site is not

It is not an engineer, and it does not know your project. It produces quantities and estimates for planning and ordering. Structural, electrical, gas and life-safety work needs a qualified person and, in most places, the approval of a building authority.

Where a figure could be relied on for something consequential, the page says what it does not cover. Please read those parts — they are the most carefully written text on the site.