Accuracy record
Where widely-published reference values are wrong, and where ours were. Every finding names the governing instrument and tells you how to check it without trusting us.
This exists because provenance is the product. Anyone can publish a number. The claim worth making is that a number was read from the instrument that creates it, on a date, and is re-checked — and that claim is only worth anything if we also publish what we got wrong.
Findings come from reading primary sources, not from comparing one aggregator against another. Where a discrepancy could not be reproduced from a public document, it is not listed. We name instruments, not vendors.
Where commonly-published values are wrong
1 finding.
Belize — public holidays do NOT all shift to Monday
Commonly published: That Belize holidays falling on a weekday other than Monday are observed on the following Monday, applied as a blanket rule. This appears on Wikipedia and is reproduced by multiple commercial holiday APIs and calendar libraries.
What the source actually says: There is no such blanket rule. Checking six consecutive official notices, the dates gazetted do not follow it: holidays are listed on their actual dates in years where the supposed rule would have moved them. The rule is a plausible-sounding generalisation that the primary source contradicts.
Instrument: Belize Statutory Instruments declaring public and bank holidays, read across six successive years rather than one
Check it yourself: Take any year's Belize public-holiday Statutory Instrument and compare the gazetted dates against what the Monday rule would predict. They diverge. One year could be an exception; six is a refuted rule.
Where we were wrong
2 errors found in data or code we had already shipped. All are fixed; each is listed with what caused it.
Colombia — we were missing a holiday created mid-year
What was wrong: Ley 2578/2026 created a public holiday on Monday 13 July 2026, effective on promulgation. We had served Colombia's 2026 calendar without it. Confirmed observed by the Cámara de Representantes. A working-day count spanning that date would have been one day out.
Instrument: Ley 2578/2026
Check it yourself: Compare our /v1/co/public-holidays against the law's promulgation date.
We returned HTTP 500 on two countries for part of a day
What was wrong: A guard added to stop working-days selling wrong answers past our calendar coverage did not account for holiday entries with a deliberately null date (ungazetted placeholders). Sorting put null last and the coverage computation threw. Trinidad and Guyana returned 500 until it was fixed. The guard itself was correct and necessary; the null-handling was not.
Instrument: (internal — commit 8087e21 through d4dc14b)
Check it yourself: POST /v1/answers/working-days for tt or gy. It now refuses or answers, never 500s.
What changed so these are caught sooner
The costliest errors we have found share one cause: a value that was correct when written and was invalidated later by a new instrument. Freshness checking cannot catch that, because the value never changed — what changed was the world. Two things now run against it:
- A weekly holiday-integrity audit that compares calendars across years, flags expired placeholder dates, and catches regional days duplicating national ones. It is what caught the Papua New Guinea additions.
- Daily source provenance — every cited source is fetched, hashed and re-checked, so if the document behind a value is amended or disappears, that surfaces rather than sitting silently. See /provenance.
Neither is sufficient. Both are better than a last_confirmed date that only says someone looked.
Found something wrong?
Send the instrument. api@3l-groupconsulting.co.za — corrections with a primary source are acted on and credited here.