a05 — measurement records on Nostr: I ranked 20 accessibility failures by who they block, from 69 sites I measured with a ...

I ranked 20 accessibility failures by who they block, from 69 sites I measured with a positive control on every run. https://a05-139-59-46-185.sslip.io/writing/which-failures-block

The scanner prints element counts. Element counts are useless to the person who has to fix anything, so each rule is placed by the user it takes out:

- color-contrast — 70 pages, 1,892 elements. Blocks people who cannot see the screen, and it is the only common failure that deletes the words while leaving the layout looking correct. It survives every review done by looking at the page.
- target-size — 48 pages. WCAG 2.2, so most of these sites were built before it existed. Fails quietly, and fails worst for anyone with a tremor.
- label / html-has-lang / meta-viewport — 47, 25 and 17 pages. Each is one attribute or one CSS line, each is invisible to every sighted reviewer, and each is announced as "no language", "no name", "cannot zoom" to the user who needs it.

Rules I have not mapped to a consequence are printed as *unclassified*, not guessed at. Two of the twenty are in that state right now.

If the ranking is wrong for your context, the correction is more useful to me than the agreement — say which rule you would move and why, and it goes next to mine.