This post is the correction-log policy and current null result. It does not imply a future reported article.
For the broader context behind this business guide, read Our Entry Bar, In Full: And Everything We Turn Down.
What this is, and who it is for
Current status: no qualifying correction entry is supplied with this edition.
That is a null result, not a claim to accuracy or rigour. The policy below remains complete and applies to any correction of a published fact.
So this is written for a reader asking a fair question: does this site actually admit mistakes, or does it quietly edit and hope nobody kept a copy? The answer cannot be demonstrated yet. What can be published now is the procedure, so that the first correction has somewhere to go and so the habit starts at page one rather than being retrofitted after the first embarrassing error.
When the log has entries, this post gets rewritten from it, item by item.
How a correction gets raised
Three routes in, and none of them requires an account or a login.
From a reader. Anything factual on this site that is wrong, stale or misattributed. An entry pointing somewhere dead, moved or unrelated. A description that misrepresents what a business does. A quote attributed to someone who did not say it or said it differently. A claim in an article not supported by the source it cites. A date that does not match the check it is supposed to represent. Report any of it at corrections.
From the operator. A business we have described can tell us the description is wrong, and it is fixed. If they want the entry gone rather than corrected, removal is faster, requires no reason, and nobody will try to talk them out of it.
From us. A re-check that finds the entry no longer matches what the page says, a second reader spotting an unsupported claim, or someone going back to a source and finding it does not say what we said it said.
What happens next is the same regardless of the route:
- The report is logged, whether or not it turns out to be right.
- If it concerns an entry, it starts a check rather than an immediate edit. A report is a prompt, not a confirmation.
- If it concerns a published claim, a person goes back to the cited source and reads it again.
- If we were wrong, the page is corrected in place with a visible note, and the correction goes in the log.
- If we were not wrong, we say so and record why, because a report that turned out to be mistaken still tells us something about how the page reads.
One thing never happens: a silent edit. Changing a published fact without a note is the specific behaviour the log exists to make impossible.
Two categories are handled differently, and both are handled by taking the thing down first and arguing afterwards. Marking a trading business as closed is the most damaging error this site can make, and it costs the operator in a way a missing entry never does. Publishing personal information that the Line says we never publish, a residential address above all, is the other. In both cases the material comes down before any discussion about whether we agree, because that order matters more than being right about it.
The corrections in this period, each with what changed
There are none.
No correction has been raised, logged, made or refused. There is no first entry to describe, no error to characterise, and no interesting story about how we caught something.
We are not going to fill this section with a hypothetical, an illustration or a worked example. A made-up correction on a page about not making things up would be an unusual choice, and an example presented carefully enough to be useful would be indistinguishable from a real one to a reader skimming.
What this section will contain, once it contains anything, is one item per logged correction with five things stated for each:
| Field | What it records |
|---|---|
| What was wrong | Stated plainly, in enough detail that a reader can judge the scale of the error. |
| What it should have said | The corrected fact. |
| How it was caught | By us, by a reader, or by the operator. Not rounded in our favour. |
| What changed | The specific edit, and whether anything else had to change with it. |
| How long it took | From report to fix. |
Which ones were caught by us, and which by a reader or an owner
Nothing has been caught by anyone, so there is no ratio, and we will not estimate one.
The reason this distinction gets its own column is worth setting out before there is any data to put in it. A log in which every error was self-caught is either a very careful publication or an incomplete log, and from the outside those two look identical. Recording who found each problem is what lets a reader tell the difference. It is also the column with the strongest incentive to fudge, which is precisely why it is published.
Externally-caught corrections are not more embarrassing than self-caught ones in any way that matters. They mean somebody read closely enough to check, which is a better outcome than nobody noticing. The report is treated the same whether it comes from a reader, an operator, or a machine consumer of the export.
The slow ones, and why they took as long as they did
None yet, and no median time to fix exists.
No turnaround claim is made. The removal page and the submission framework use the same rule: an implementation may state a median only when it is computed from its dated log, including misses.
What can be said now is which things will not be allowed to be slow. The two harm categories above come down first, before assessment. Everything else runs at the speed a person can do it, and we would rather publish an uncomfortable duration later than an invented one now.
There is one honest complication worth naming in advance. Corrections that require a re-check are inherently slower than corrections to an article, because a re-check means running the method rather than editing a sentence. When the log shows slow items, that will often be why, and saying so in advance means we cannot use it later as an excuse invented after the fact.
What changed in the method as a result
Nothing, because nothing has been corrected.
The distinction matters and it is easy to blur. Errors go in the correction log. Changes to the procedure that produces dates, and changes to the inclusion criteria, go in the method changelog, which currently records one version: the first published one. A method version is not an error, and filing one as a correction would inflate the log with things that were never mistakes.
Where the two meet is the interesting part. A correction that reveals a hole in the procedure should produce a method version, and when that happens the changelog entry will say which correction prompted it. That link is the thing that makes a changelog honest, because a procedure that never changes in response to a real error is a procedure nobody is testing.
How to raise one
Tell us at corrections. Include the page, what you think is wrong, and what you think it should say. Nothing else is needed, and there is no form gate, account or verification step in front of it.
A few practical points:
- You will not be named unless you ask to be. The log records whether an error was caught internally or externally, not who found it.
- Being wrong is fine. Reports are logged either way, and a mistaken report usually still tells us something about how a page is being read.
- If you are the operator and you want the entry gone, removal is the faster route. It is free, unconditional, and requires no reason.
- If something about your entry has changed rather than being wrong from the start, tell us what changed and we will update or re-check it.
- There is no charge, for corrections or for anything else on this site, and no payment of any kind moves a report up a queue.
In Plain English
This edition supplies no correction row, ratio, turnaround median, or method-change result. What it supplies is the complete procedure: reports are logged whether or not they are right, claims are checked rather than instantly edited, wrongly marked closures and exposed personal information come down before discussion, and no published fact changes silently. No future rewrite is promised or counted as evidence.
Report something we got wrong. It costs nothing, you will not be named, and a mistaken report is still worth sending.