Where the formulas come from
Each calculator implements a published formula, and the page names it where it has a name. Health and medical equations are taken from the original paper or the publishing body — Mifflin-St Jeor from the 1990 study, the US Navy body-fat equations from Hodgdon and Beckett, gestational dating from Naegele's rule, growth percentiles from the CDC and WHO data files themselves rather than a secondary summary. Unit conversions use exact international definitions, not rounded approximations: an inch is 0.0254 metres exactly, a pound is 0.45359237 kilograms exactly.
234 of 234 calculators cite at least one source on the page. The ones that do not are pages where a citation would be theatre rather than evidence — a tip split or a movie runtime total is arithmetic, not a claim about the world.
How they are tested
Every calculator ships with automated tests that run on each build. These are not smoke tests; they assert specific values:
- Reference values. Known answers from textbooks or the source document — a $200,000 mortgage at 6% over 30 years must produce $1,199.10 a month.
- Property tests. Facts that must hold for every input: converting a value and converting it back returns the original, the parts of a breakdown sum to the total, a closed-form answer matches a month-by-month simulation.
- Guard tests. Bad input must produce a clear message or nothing at all — never a wrong-looking number. Division by zero, impossible geometry, and values crafted into a shared link are all covered.
- Sitewide gates. Every calculator is run with edge values in every numeric field to prove it cannot display
NaN, and each one must compute with its defaults before it can be published at all.
Why the article text is checked too
A calculator can be correct while the explanation around it is wrong, and readers reasonably trust both. So every figure that appears in an explanation or worked example is recomputed against the calculator's own output. This check has caught genuine contradictions — a materials estimate whose text said 74 bags where the tool correctly said 75, a projection whose written breakdown dropped the opening balance.
The independent audit pass
After a calculator is built, it is audited by someone who did not build it, briefed to find errors rather than to confirm the work: to re-derive the maths independently, re-fetch the cited sources, and check whether stated conventions still hold. Findings from that pass have included a regulation that had been superseded, a constant that disagreed with the paper cited beside it, and reference-table values that did not match the table they were credited to.
Known limitations
Results are estimates. Published formulas are population averages, and your own case can sit well away from the average — this is especially true of body composition, energy expenditure, alcohol metabolism and pregnancy dating, where individual variation is large and a professional has information a web page does not.
Material quantities, appliance ratings and coverage figures are typical values that vary by product; the page says so where it uses them, and your supplier's figure beats ours. Rules set by governments and lenders change, so you enter the current rate rather than trusting a number frozen at publication time. Where a simplified model is used instead of the authoritative one, the page says which and why — a flat sizing rule of thumb is labelled as exactly that.
Corrections
If something here is wrong, tell us and it gets fixed. Email hello@calcduck.com with the page, the value you expected and your source. Corrections are handled by the CalcDuck Review Team.