A learning platform's conformance report describes the platform as the vendor built and tested it. It does not describe how your team configured it, and it plainly does not describe the four hundred courses your instructors uploaded. Those are three different objects with three different owners, and treating them as one is how a procurement conversation ends with everybody surprised.
The split matters commercially as well as technically. A finding in the platform is the vendor's to fix and yours to escalate. A finding in the configuration is a settings change somebody on your side can make this afternoon. A finding in a course is a content problem that belongs to whoever wrote the course, and there may be hundreds of them behind one report line.
The Standard That Names the Gap Nobody Audits
WCAG covers what a reader receives. A learning platform is also a tool your instructors publish through, and there is a separate W3C standard for that, which is why an editor can pass a WCAG audit and still make accessible courses impossible to write.
ATAG 2.0 has been a W3C Recommendation since September 2015, and it comes in two parts. Part A makes the authoring tool's own interface accessible to authors with disabilities. Part B is the one owners care about, because it covers whether the tool supports all authors in producing accessible content. Its definition of an authoring tool is broad enough to catch anything used to create or modify web content for other people, so a course builder, a quiz builder and a discussion board are all inside it.
One criterion tells you what Part B feels like in practice. B.2.3.1 says that if the tool lets authors add non-text content, authors must be able to edit the text alternative associated with it. If your instructors cannot type alt text into the image dialog, no amount of instructor training fixes that, and no WCAG audit of the editor's interface will report it either. ATAG and UAAG covers the rest of the family.
What we do and do not assess
We audit against WCAG. We do not carry out ATAG assessments, and this page names the standard so you can tell which questions a WCAG audit answers and which it does not. A WCAG evaluation of an authoring interface tells you whether an instructor with a disability can operate the editor. It does not tell you whether the editor helps anybody produce accessible output, and nobody should let one be sold as the other.
The Three-Layer Matrix
This is the artifact to build before anyone quotes for anything. Four columns, three layers, and the fourth column is the one that stops arguments later.
| Layer | Who owns a fix | What evidence usually exists | What that evidence cannot say |
|---|---|---|---|
| Platform interface: navigation, dashboard, course player, quiz engine, discussion board, gradebook | The vendor | A published conformance report, scoped to the release and configuration the vendor tested | Whether your release, your theme and your enabled modules behave the same way |
| Your configuration: enabled modules, theme, roles and permissions, timer defaults, integrations, single sign-on | Your administrators | Usually nothing, because nobody evaluates a settings screen | Anything at all, until somebody tests the instance rather than the product |
| Course content: pages, uploaded documents, embedded video, images, quizzes, packaged modules from a publisher | Whoever authored each course | A sample, at best, and often only the flagship course | Whether the other courses in the catalogue resemble the sampled ones |
Every argument about who fixes what resolves in column two. Every argument about whether a report covers something resolves in column four. Fill this in with your real module list and the scoping conversation gets shorter, not longer.
Course Content Is a Sample, and the Sample Has to Be Honest
Nobody audits four hundred courses. What a real evaluation does is take a structured sample that reflects the shapes in your catalogue, then add a small random set on top to check the structured one was not flattering. WCAG-EM puts the random set at a tenth of the structured set, added rather than carved out, and if the random courses turn up problems the structured ones did not, the structured sample was wrong and you widen it.
The shapes worth covering in a course sample are more varied than most people expect. A video-led course and a document-led course fail differently. A course with a graded quiz and a course with a reflective journal exercise a different half of the platform. A publisher's packaged module is somebody else's content sitting inside your player, and a course authored in 2019 was probably built with an older editor.
- A media-heavy course, which exercises the player, the captions and the transcript route.
- A document-heavy course, which exercises whatever your instructors uploaded, usually as PDFs. PDF accessibility is a separate evaluation with a separate unit of work, and a PDF that never got tagged is not a platform finding.
- A course with a graded, timed assessment, which is where the timing argument below actually lands.
- A course with a discussion requirement, because participation grades hang off a component nobody demos.
- A packaged module from a publisher, which tests the integration as much as the content.
- An old course nobody has touched, because it was authored in a version of the editor that no longer exists.
Which Timers Can Argue the Exception, and Which Cannot
Timers are where a course platform audit gets interesting, because the rule has more give in it than people assume and less than platforms hope.
Timing Adjustable is satisfied by any one of six things. Three are actions you can take, turning the limit off, letting the reader adjust it, or warning and letting them extend. Three are exceptions, a real-time event, an essential limit, and a limit longer than 20 hours. The extend route has exact numbers attached, and they are at least 20 seconds' warning, a simple action to extend, and at least ten extensions available.
The essential exception is the one an assessment reaches for, and it is genuinely available, because extra time really can change what a test measures. But essential is a defined term with two halves. Removing the limit would have to fundamentally change the information or functionality, and the same result would have to be unachievable another conforming way. A graded assessment can argue both. The clock on the enrolment page cannot, and it does not inherit the argument by living on the same platform.
| Timer | Route to satisfy 2.2.1 | What to record |
|---|---|---|
| Graded assessment countdown | Essential exception, argued on the assessment's purpose | Who decides the limit, and whether an extended-time configuration exists for approved learners |
| Session inactivity timeout | Warn and extend, or adjust | The warning interval, the extension action, and how many extensions are allowed |
| Enrolment or payment hold | Warn and extend, usually | Whether the hold is inventory-driven, which is an argument to make rather than assume |
| Auto-advancing course slide | Turn off or adjust, and Pause, Stop, Hide applies as well | Whether a control exists and whether it is reachable by keyboard |
| Live session countdown | Real-time event exception, where the session genuinely is one | That the recording route afterwards is not itself timed |
Enrolment and Submission Are Processes, So They Pass or Fail Together
WCAG's third conformance requirement is the one people underestimate. Where a page is one of a series presenting a process, every page in that series has to conform for any page in it to conform. Enrolling is a process. Submitting an assessment is a process. Getting a certificate issued is a process.
That has a blunt consequence for reporting. A learning platform whose course pages are clean and whose payment step is not does not have clean course pages, as far as the process is concerned. So the sample has to follow whole journeys rather than sitting on representative pages, and the report has to say which journeys it followed to the end.
What to Ask Your Vendor, and What to Ask Yourselves
- Which release and which configuration does your report cover. Not whether the platform is accessible. The report's own scope statement is the only thing that answers this, and it is usually two sentences near the top.
- Which modules were in scope. A report covering the core platform may say nothing about the module you bought last year.
- What settings affect accessibility. If a default fails and a setting fixes it, that is the cheapest finding in the whole engagement and it only exists if somebody tells you.
- Whether the editor lets instructors add text alternatives, heading structure and table headers. That is Part B territory, it will not be in a WCAG report, and it decides what every course after today looks like.
- Who fixes course content. This is the question that decides the size of your remediation problem, and it is an internal one. A platform vendor cannot fix a course they did not write.
One honest limit
We audit and never repair, so we evaluate the course experience against WCAG on the courses and configuration you put in scope and stop there. We do not fix courses, recaption video libraries or assess authoring tools against ATAG, and where a finding belongs to your platform vendor rather than to you, the report says so and keeps it in its own pile. If your catalogue is large, the sample is the conversation, so start it at contact or read what the LMS audit covers.