When this publication gets something wrong, it says so on the thing it got wrong — dated, in place, where the person who read the error is most likely to read the fix. A page is never quietly edited to make the mistake disappear.
The rule
Corrections are published, dated, and visible on the corrected item — never silently edited.
That sentence comes from section 5 of the editorial independence policy, which established this policy because no corrections policy existed anywhere in the operation before it. It has three parts, and all three are load-bearing:
- Published. A correction is itself a piece of published material. It goes in the log on this page.
- Dated. Every correction carries the date it was issued, so a reader can tell how long the error stood.
- Visible on the corrected item. The correction appears with the story, the table, or the figure that was wrong — not only here. A log nobody visits is a way of technically disclosing something.
The thing all three exist to prevent is the silent edit: changing the words so that the record now shows a publication that was never wrong. A publication that edits its own past has no past.
What counts as a correction
- A figure that is wrong — the wrong number, the wrong period, the wrong company, or a number attached to a source that does not state it.
- A name, date, or quotation that is wrong.
- A characterisation the source does not support — a conclusion drawn past what the record actually says.
- A link that cites the wrong document.
What is not a correction
Two cases that look like errors and are not, stated so the distinction is not made up later to avoid an awkward entry:
- A figure that changed at the source. When a company restates a number in a later filing, this publication's earlier reading of the earlier filing was not wrong. The page follows the record; a restatement is news about the company, not an error here.
- A figure that is missing. An em-dash is not an error. It is this publication declining to guess, which is the intended behaviour and is explained on the about page.
Anything that is genuinely arguable gets an entry. The bias runs toward logging.
One honest complication
The front page is regenerated from the underlying public records on a schedule, so what it shows can change between two readings without anyone touching a word of it — because the filings changed, or because a figure that was an em-dash became available. That is the design, and it is why the log matters more here than it would for a publication whose pages sit still: a correction has to be a deliberate, recorded act, or it is indistinguishable from the page simply being rebuilt. Nothing in the regeneration removes an entry from this log.
How to report an error
There is no public channel for reporting an error yet, and this page is not going to pretend otherwise. No inbox, no form, no address. That is a gap, it is named here rather than buried, and it will be filled by putting a real address on this line — not by adding a form that goes nowhere.
Until it is filled, corrections happen the only way they currently can: someone working on the publication finds the error, or a reader who already has a way to reach the operation reports it. Every correction that results is logged below, whatever route it arrived by.
The log
| Date | What was corrected | What was wrong | What it says now | Source |
|---|---|---|---|---|
| No corrections issued yet. | ||||
This log is empty because no correction has been issued, not because entries have been removed. It is generated with the rest of the site from a list held in the publication's source, so an entry is added by a change with an author and a date on it, and removing one would be equally visible. Entries are never deleted.