Corrections

How to report an error in a number, a claim or a label — and how we log the fix.

Updated October 19, 2026

We publish numbers, and numbers can be wrong. If a figure on this site does not match its source, tell us and we will check it against the pipeline. Corrections are part of the editorial record, not a PR exercise.

How to report a correction

Email [email protected] with:

  • the page URL,
  • the number or sentence you believe is wrong,
  • the number you believe is correct, and
  • the source or calculation behind it, if you have one.

A screenshot is helpful but not required. If the issue is a broken link, a missing page or a typo, the same address works.

What we correct

  • Data errors — a rate, count, ratio or sample size that does not match the underlying aggregate.
  • Labelling errors — a claim described as a measured failure rate when it is a mention rate, or a comparison that mixes two populations.
  • Attribution errors — a quote, study or dataset credited to the wrong source.
  • Broken links and outdated references — including links to pages we have retired.
  • Typos and formatting — fixed silently; no correction note is needed.

What we do not change

We do not remove a finding because a brand, seller or reader disagrees with the interpretation. If the number is right and the method is stated, the finding stays; we will publish a response or a caveat if the disagreement is substantive. We do not accept payment to change a score, a ranking or a finding.

How we log corrections

  • Material corrections (a number, a claim, a conclusion) are noted at the bottom of the affected page with the date and what changed, and are listed on the changelog .
  • Non-material edits (typos, formatting, link fixes) are made in place; the page’s lastmod date changes.
  • Retractions are kept visible: the page is updated with a note explaining what was withdrawn and why, rather than silently deleted.

Timelines

We aim to acknowledge a correction request within five business days. A data correction takes longer if it requires re-running the aggregation pipeline; we will tell you the expected date rather than guess.

Share