An accessibility audit is a sample, a method and a report. The sample decides what it can honestly say, the method decides whether the findings are reproducible, and the report decides whether anybody can act on it. Most quotes describe the report and leave the first two vague, which is how two audits of the same site arrive at very different prices and very different page counts. On the worked example a 240-page site sampled at 8% is 19 pages opened, 11 hours of manual testing, $3,480 for the audit and $1,392 for the retest.
The sample is the audit
An auditor cannot open every page and should not pretend to. A defensible sample covers every template, every interaction type (form, search, checkout, media player, modal) and a few pages chosen at random. On the worked example nineteen sampled pages across nine templates is two pages a template, which is the minimum that shows whether a failure lives in the template or in the content.
The method decides whether findings are reproducible
Ask which WCAG version and level, which browser and screen reader pairs were used, and whether testing included keyboard-only and 200% zoom. Findings without those details cannot be reproduced by your developers, and a finding a developer cannot reproduce is a finding that never gets fixed.
What it costs, and what the retest is for
Manual hours plus write-up, times the rate: 11 hours of testing plus 12 hours of write-up at $150 is $3,480, of which $1,800 is the write-up alone. The retest at 40% is $1,392 and is the part that proves the fixes landed. Budget both at the start, because an audit with no retest produces a list nobody ever closes.
Questions people ask about accessibility audits
How long does an audit take?
Usually two to three weeks from sample agreement to report, most of which is scheduling rather than testing.
Should we scan first?
Yes. Clearing the machine-findable failures before the audit means you are paying an auditor only for the judgement a person has to make.
How many pages should be sampled?
Enough to cover every template and every interaction type. Percentage alone is a poor guide: a 2,000-page site on six templates needs fewer sampled pages than a 200-page site on thirty.