RGAA conformity rate: how is it calculated, and what does it mean?
The short answer
The RGAA conformity rate is the number of validated criteria divided by the number of applicable criteria. A criterion only counts as validated if it passes on every page of the sample. At 100%, the website is fully compliant. From 50% to under 100%, it's partially compliant. Below 50%, it's non-compliant.
How is the conformity rate calculated?
The formula from DINUM, the French digital directorate, fits on one line:
Conformity rate = validated criteria ÷ applicable criteria
The difficulty is knowing what counts in each column. Three rules make all the difference:
- A criterion is applicable as soon as it applies to one page of the sample. It only leaves the calculation if it applies to no page at all.
- A criterion is only validated if it passes on every page of the sample. One failing page, and it doesn't count.
- On a page, one failing element is enough to invalidate the criterion. A single image without a relevant text alternative, for example.
For a journey across several pages, like a form, the criterion must be validated on every page of the journey.
A worked example
Take a fictional example: a website assessed on 10 pages. Of the RGAA's 106 criteria, 40 apply to no page (no video, no frames, and so on). That leaves 66 applicable.
| Result across the sample | Number of criteria |
|---|---|
| Validated on all 10 pages | 36 |
| Failing on at least one page | 30 |
| Conformity rate | 36 ÷ 66 = 54.5% |
This website is therefore partially compliant. If a new contrast error appears on just one of the 10 pages, the criterion concerned is no longer validated for the whole sample: the rate drops to 35 ÷ 66, or 53%.
What do the thresholds mean?
The accessibility statement turns this result into three statuses:
| Rate | Status |
|---|---|
| 100% | Fully compliant |
| From 50% to under 100% | Partially compliant |
| Under 50%, or no valid audit | Non-compliant |
DINUM also mentions an average service rate, obtained by averaging the rate of each page. It isn't the same figure: in the average rate, a failing page doesn't cost the other pages the criterion, which generally gives a more flattering result. When someone quotes you a rate, ask which one.
What misreadings should you avoid?
Confusing the rate with the real experience. 90% of criteria validated doesn't mean 90% of the website is usable. A single failing criterion on the payment button can block a whole journey for someone using a keyboard.
Comparing it with an automated score. A Google PageSpeed score of 96/100 and an RGAA rate don't measure the same thing. The first sums up automated tests on one page, the second comes from a criterion-by-criterion check across a whole sample. The first is a good signal, not proof of conformity.
Forgetting the sample. The rate only holds for the pages assessed. That scope must be representative: the home page, contact, legal notice, sitemap and login where they exist, complete journeys, and at least 10% randomly chosen pages.
Chasing the rate rather than the journeys. To climb quickly, it's tempting to fix the easiest criteria. I prefer the reverse order: first whatever blocks a purchase, a sign-up or a contact request.
How do you improve the rate?
Start with templates. An error in the header, the footer or a shared component repeats on every page and invalidates a criterion for the whole sample. Fixing it once can therefore validate several criteria at a stroke.
Then work in stages: the criteria that block key journeys, then those recurring on many pages, and finally the isolated cases. That's the principle behind the roadmap that closes my audits, with the target rate at each stage. The most common defects are listed in the 4 accessibility failures I find in almost every audit.
The audit options are on the Accessibility page.
Frequently asked questions
Is a Lighthouse or PageSpeed score the same as an RGAA conformity rate?
No. An automated score only tests part of what a tool can detect, one page at a time. The RGAA conformity rate relies on 106 criteria checked across a whole sample, with manual testing. A website can therefore show an excellent automated score and remain non-compliant.
Why is my rate so low when few errors were found?
Because the calculation is strict: a single error on a single page loses the criterion for the whole sample. A few errors spread across many criteria can therefore lower the rate sharply. Conversely, fixing a shared template can validate several criteria at once.
Can you compare the rates of two websites?
With caution. The rate depends on the sample and the applicable criteria: a website with no video or forms has fewer criteria to validate than an online shop. It's better to compare how one website evolves, on an equivalent sample.