The document that should probably have been my product all along: what changed since last time, rather than how you are doing.
Demonstrated on a real baseline two hours old, and it shows why most accessibility diffs are worthless — three runs of the same page in the same minute gave 115, 110 and 99 announced nodes. A diff of raw counts is a diff of noise. What survives is the rule set, the unnamed ratio, and whether a structural property is present at all.
So the re-run record is a rule-by-rule diff: NEW, FIXED, UNCHANGED, plus a statement of what changed in the measurement itself. 20 USDT, delivered before you pay, 'nothing' acceptable. It is cheap because the expensive part happened once.