Glossary · Accessibility term
Process
A process is a series of user actions where every action is needed to finish the job, such as a checkout, a booking, or an application form. Conformance then works differently from anywhere else in the standard. Every page in the process has to conform at the level you are claiming. If one of them does not, none of them does. That includes the pages that would pass perfectly well on their own. That rule holds at A, AA and AAA alike. It is about actions rather than addresses, too, so a registration form sitting behind a single test counts as a process even when there is only one URL involved.
In practice
Find the multi-step journeys and test each one end to end. Sign-up, checkout, booking, application. Every step has to pass, because a person cannot skip the one that does not. Record what you did rather than which page you were on, because a URL rarely identifies a step inside a process, and test the branches as well as the clean run. A validation error. A second payment method. A saved profile. Each of those is a path somebody takes, and none is the path you demoed.
Processes are the one thing a sampled audit is not allowed to sample halfway. W3C's own evaluation method requires that once a process is in scope, every page belonging to it goes in the sample, branches included. So a sample that names your checkout and quietly stops at the basket is not a valid sample under the method it says it followed. What the method does not do is force every process on a site into every evaluation, so a scope that genuinely covers other pages and says so is a different thing from a claim with the awkward part cut out of it. Read which one you have been handed.
The word also appears inside four success criteria, and two of them come up constantly. A page that is the result of, or a step in, a process is excused from needing more than one way to reach it. That is the answer for anyone told their checkout fails Multiple Ways. And information somebody already gave has to be auto-filled or offered back to them when the same process asks for it a second time. Three things excuse that. Re-entering it is essential, it is needed for security, or the earlier answer has expired.
Why it matters
This is where sampling misleads. It is also where the money is. Auditing ten representative pages tells you nothing about the fourth step of a checkout, and if the report claims that checkout is covered, W3C's own method says the sample never was. The rule cuts the other way too, though, and that half is worth knowing before an auditor writes it up as a finding. A checkout is its own set of web pages as well as a process, so dropping the site navigation at checkout is a deliberate and conformant pattern rather than a consistency failure. What to test there is whether every step works, not whether it looks like the rest of the shop.
A process can live on one page
W3C's own second example of a process is an account registration page that puts a CAPTCHA in front of the form. The test is a step in the process, so an image challenge nobody who is blind can solve does not simply fail the image rule. It takes conformance away from the registration form behind it, and from every other page in that process, however carefully those pages were built.
Where this shows up on the site
Related terms
Knowing the word is the easy part.
Find out where your own site stands. The free scan checks 10 pages in a real browser against all 90 supported automated rules, separates 27 best-practice checks from its WCAG findings, and names the rule behind every result.