Why we publish our corrections

Our corrections log records material errors — factual claims that turned out to be wrong, attributions to the wrong source, numerical errors that change a rubric score, missing context that changed a story's framing. We publish the original error, the correction, and the date.

This is the only meaningful evidence we can give that our editorial process catches mistakes, that we take them seriously when caught, and that we are not silently revising the historical record. Without a published log, there is no way for a reader to distinguish 'a publication that doesn't make errors' from 'a publication that doesn't acknowledge errors.'

A correction is required when: a factual claim materially shapes a conclusion; a numerical value affects a ranking or score; an attribution misnames a source; or an editorial decision misframes a topic in a way that materially changes a reader's understanding.

Minor typographic errors, formatting fixes, and style updates are not logged as corrections. They are silently edited. The distinction is whether a reasonable reader, having read both versions, would feel that something material about the article had changed.

A correction is also a process artifact. When we issue one, we review where in the editorial pipeline the error escaped detection — author, copy edit, medical review, fact check — and adjust that pipeline if a category of error recurs.

The most common category we've adjusted on, since launch, is provider price changes that we didn't catch within our refresh cadence. The fix has been a quarterly explicit re-verification of priceLabel and priceMonthly across the provider list. The corrections log records the change; the operational fix records the lesson.