Every figure here can be checked. Some of them will still be wrong.

What we do when a number is wrong

1. The figure is fixed, or it is removed. If we can compute the right number from a filing we can name, the page prints the right number. If we cannot, the page prints nothing rather than a second guess. A missing figure reads n/a here, never zero and never an estimate wearing a decimal point.

2. The page carries a visible note. The byline on a corrected page prints the date it changed and one line saying what changed, beside the date it was published. It is in the body text at the top of the page, not in a log somewhere else, so a reader who arrives at the corrected page sees that it was corrected without going looking.

3. The modified date moves, and only then. Every page here declares two dates: the day it was published and the day its text last changed. The second one moves when the words or the arithmetic move, and on no other day. It never moves because a build ran.

   This is enforced rather than remembered. A page whose modified date is later than its published date must carry the note saying what changed, and the build fails if it does not. That is one mechanism, not two: the note printed beside the date, the date in the page's structured data, and the entry on this page are the same event described in three places.

4. It is listed below. Every correction to a published figure gets a row: the day the fix went live, the page, what the number was and what it is now, and how it got there. The last of those is the part worth reading, and it is the part a log of dates and diffs leaves out.

What counts as a correction

A published figure that was wrong, a source that was named and was not the source, or a claim about what this product does that it does not do. Those get a row.

Rewriting a paragraph, adding a section, or fixing a typo does not. Those move the modified date and print the note, because the text changed, but nothing was wrong. A corrections list that fills up with copy edits is a corrections list nobody reads, and the point of this one is that a row on it means something.

A figure that changes because a company restated its accounts is not a correction either. That is the filing changing, not us being wrong, and the methodology describes how a restatement is picked up.

Corrections made

None. Nothing published here has been found wrong and corrected since this page went live on 2026-09-07.

That is a short record rather than a good one. This site is days old and has published a handful of pages, so an empty list here is mostly a statement about how little has been published, and it is printed as one rather than as a claim about accuracy. The list is worth reading in a year. It is worth publishing now because the alternative is starting it on the day it first has something to hide.

Tell us about one

Send the page, the figure and the filing you checked it against, to hello@valuescreener.ai. Naming the statement and the fiscal year you read is the fastest possible version of this, because it is what has to be checked either way.

A number that turns out to be right gets an answer explaining why, not silence. Where two sources disagree and both are defensible, the page says which one it uses and what the other one says, which is a case the definitions already deal with at length.