How we test

Every tool is two separate pieces: the arithmetic, and the page around it. The arithmetic has no access to the page and is run by an automated test suite against a list of worked examples. The site's build runs those tests first. If any example fails, nothing is published.

The same list of examples is printed on each tool's How it works page, so you can see exactly what it has been checked against and work one through by hand if you want to.

Published means the expected answer is printed in the cited source: a code table, a manufacturer's instruction sheet, or an author's worked example. We copy the inputs and the answer.

By hand means the source states a rule but no worked answer for that case, so the expected value was worked out by hand from the rule. These catch arithmetic and edge-case mistakes, but they cannot catch a misreading of the rule itself, which is why they are labelled differently.

Tests show the calculator follows the cited rule. They do not show the rule applies to your job. Building codes are adopted in different editions with local amendments, manufacturers' instructions can override general rules, and engineered products have their own tables. Each tool says which source and edition it uses.