Reviews
Part of People by heritage in one sourced place in 2027
People by heritage updates beyond the obvious
People by heritage updates for 2027: reference panels, archive openings and category wording move, and a review checklist for pages that quote them.
Pages about origins go stale in an unusual way. The records do not change. What changes is the apparatus around them: which archives are searchable, how a company's comparison set is built, and what words a statistical agency puts on a form.
This page is a review procedure for that apparatus, written so a re-check takes an afternoon rather than a rewrite.
What to take away
- A commercial ancestry estimate can change without the sample changing, because the comparison set changed.
- Archive openings move what is findable, which looks like a change in what is true.
- Category wording on official forms is revised between editions, and old figures do not convert.
- Nothing on this site is refreshed by adding a living person's origin, because that field is not kept here.
The four things that actually move
| What moves | Who moves it | What it invalidates |
|---|---|---|
| Reference panels | Testing companies | Any stated percentage from a previous run |
| Archive access | Archives and record agencies | Statements that a record cannot be found |
| Category wording | Statistical agencies | Comparisons across census or survey editions |
| Access law | Legislatures and courts | Statements about what is obtainable, and by whom |
Reference panels and the numbers that follow them
A consumer ancestry estimate is a comparison. The sample is measured against panels assembled from present-day people who reported particular origins, and the output is similarity to those panels. Change the panels and the output changes with no new information about the person. Public health guidance on direct-to-consumer genetic testing sets out the same point about how these results should be read. A people by heritage facts page should state which panel version produced any percentage it quotes.
Two review consequences follow. A percentage quoted on a page needs the company, the panel version, and the date, or it cannot be re-checked at all. And a page that describes such a result as an identity rather than as an estimate is wrong on the day it is published, not just later.
Full siblings receive different samplings of the same ancestors' material, so their estimates differ even with identical documented ancestry. If a page treats a percentage as a property of a family, that is the check that catches it.
Archive openings
Records become findable in bursts: a series is digitized, an index is published, a restriction expires. Each event changes what a search returns without changing what happened.
On review, re-run the searches that produced a stated absence, and re-read the scope notes for the series involved. National agencies publish the state of their holdings for each research area, including census records, and their notes are the fastest way to see what has moved. Record the search date next to the result so the next reviewer knows how old the absence is.
Category wording
Official categories are answer options, and the options are revised. Names are added, retired, merged, and split; some editions allow one selection and later ones allow several; some rounds were recorded by an enumerator and others by the household. A figure from one edition does not convert into the next, and a percentage change across an edition boundary is usually a change in the question.
On review, pull the actual questionnaire and coding instructions for each year involved before touching the numbers. If a page compares two editions without naming the wording change, the comparison comes out and a note goes in.
The standing review checklist
- Does every quoted estimate name its company, panel version, and date?
- Does every stated absence carry the series, index, and search date?
- Does every cross-edition comparison name the wording change between them?
- Does any sentence describe a living person's origin, ancestry, religion, or nationality? If so, it comes out.
- Does any table sort places or communities by counts of people? If so, it comes out.
Items four and five are not discretionary here, and the reasoning behind both is set out on the heritage pillar. Related material sits on the country pillar for jurisdictions, the century pillar for periods, and the maps pillar for anything drawn as a map.
Common questions
A reader sends a new test result that contradicts a published page. What now?
Nothing changes unless the page made a claim it should not have. A new estimate is a new comparison against a new panel, not a correction of a fact.
How often should the review run?
On events. A panel update, an archive release, or a new census edition is a trigger. A date in the calendar with none of those is not.
Should old figures be deleted when a new edition changes the categories?
No. Keep them, labeled with the edition and wording that produced them. Deleting them makes the discontinuity invisible, which is the error the label was preventing.
What if a subject asks for a change to how their background is described?
Where the page quotes their own public statement, update it to their current one and date both. Where the page asserted something they did not say, that sentence should not have been there, and the fix is removal rather than rewording.
