Skip to main content
WCAGrules
Quick navigation

Free tools · Helper

Accessibility audit scope builder

Most audit briefs ask for a page count, which tells a supplier nothing and is why two quotes for the same site come back ten times apart. Name your views, your journeys and the states they run in, and this writes the scope instead, with the sample arithmetic and the sentence about what a sample cannot answer.

Which common views does the product have?

One of each kind, not one of each page. Two pages built from the same template are one view.

  • Where most first visits land.

  • Your commonest template, whatever it is called.

  • Repeated structure, and search has its own patterns.

  • Captions, transcripts and alt text live here.

  • Contact, feedback, a single-step signup.

  • The download route, separately from the files themselves.

  • How somebody gets anywhere.

  • Content that appears without a page load.

  • The gate in front of everything behind it.

  • Where a user changes how the product behaves.

  • The pages somebody reaches for when the rest has failed them.

  • The 404, and whatever the server shows when it breaks.

Which journeys have to be finished end to end?

A page inside a process only conforms if every page in that process does, so a journey enters the sample whole. Count the steps in the path somebody takes when nothing goes wrong, then the branches that matter, like paying a different way or fixing a rejected field.

No journeys yet. Most sites have between two and five that matter: sign up, sign in, search and buy, book, get in touch.

Which states and variations exist?

These do not multiply the sample. They say what every sample has to be looked at in, and they are the half of a scope that briefs usually leave silent.

  • Part of the full page rather than an extra sample. Each variation the page presents automatically has to conform for the page to conform.

  • Two views. Whichever one you skip is untested, and it is usually the signed-in one.

  • The state where a journey actually breaks, and the one a demo never reaches.

  • A separate view, not an automatic variation, so it needs its own coverage rather than inheriting the default theme's result.

  • Each locale is its own content. A tested locale says nothing about an untested one.

  • Name which variant is in scope. An unevaluated branch is untested, which is a different status from working.

What should it be tested against?

WCAG names no combinations, so the evaluator sets a baseline with you before testing. Leave this empty and the scope will say it is still open, which is better than leaving it unsaid.

Your scope so far

Nothing selected yet. Tick a view or add a journey on the left and the scope builds here, with the sample arithmetic and the limit that goes with it.

What it checks

  • Your common views, counted once each

    One of each kind rather than one of each page, because two pages built from the same template are one view. The list follows the sample GOV.UK asks an auditor to cover.

  • Your journeys, whole

    A page inside a process only conforms if every page in that process does, so a journey enters the sample complete: the default path, plus the branches that matter.

  • The sample arithmetic

    A structured sample plus a random set at 10 percent of it, added on top rather than taken out. That is W3C's method. Rounding up to the next whole sample is ours, and the page says so.

  • The states every sample runs in

    Signed in and out, error states, themes, locales and variants. These do not multiply the count. They say what each sample has to be looked at in, which is the half of a scope most briefs leave silent.

  • Whether a baseline exists yet

    WCAG names no combination of browser and assistive technology, so an evaluator agrees one with you before testing. If you have not, the scope says so rather than staying quiet about it.

What it cannot tell you

  • Whether your sample is representative. That gets tested during the evaluation, by comparing the random set against the structured one, and it can send an evaluator back to pick more. This runs before any of that.
  • How many views a site should have. No standard sets a number and neither does this. The count you see is your own selection added up.
  • Anything about how accessible any of it is. A scope is a plan. There is no score here, no percentage and no verdict, and W3C's own methodology argues against aggregated scores in plain terms.
  • What a sample can claim. It cannot support a conformance claim for a whole site, because it is always possible that a page nobody opened carries an error. No sample size answers that, so the scope carries the limit with it.

Where to go next

Go somewhere useful

Find tools, resources and your workspace.

29 destinations

↑ ↓ navigate · Esc closeSearch the full site