Research
Findings about accessibility conformance, published in full - the method, the numbers, and the findings that did not survive checking.
How These Are Published
Each study says what was tested, when, and how, and publishes the result whether or not it flatters anyone. Where software produced a finding, a person checked it before it counted, and the ones that failed that check are listed next to the ones that passed.
Every figure carries the date it was gathered, because a number about the web goes stale. If a study is re-run, the page says so and the date moves.
The ACR Availability Finding
Published
We went looking for a published accessibility conformance report at 107 recognizable SaaS vendors across six verticals, all on one date. Thirty publish one openly. The rest either put a form, a login, an NDA or a sales conversation in front of it, or had nothing we could reach from their public pages. Open publication runs from 58% in the most open vertical down to 5% in the least.
Scope: Each vendor landed in exactly one of four buckets, and the four sum to 107. No vendor is named anywhere on the page.
A Worked Section 508 Evaluation
Published
Automated analysis of a public federal homepage flagged issues across eight success criteria. Checking every one by hand against the live page left 10 confirmed defects and threw away 15, each listed with the reason it did not survive. The discarded ones are the point: they were plausible and specific and wrong, and every one of them would have reached a client as a defect to fix if nobody had looked.
Scope: One page, at 1440x900, in Chromium, on July 31, 2026. We verified the failures, not the passes, so nothing in it is a conformance claim about the site we tested.
Related Reading
The same evidence, pointed at the decisions a buyer actually has to make.