Conformance Drift: Why Your Accessibility Score Decays Between Audits and How to Catch It
2026-10-18
An audit is a photograph. It shows conformance on one day, for the pages sampled, in the state they were in. The day after, the site begins changing again: a marketing landing page ships from a new template, a checkout widget updates itself, an editor publishes forty product pages from a spreadsheet. Each change is small. Together they are why last year's clean audit and this year's messy one are both accurate.
Programs that report a single annual score cannot see drift, so they cannot manage it. The score falls off a cliff once a year and everyone acts surprised. Treating conformance as a trend, measured continuously and attributed to releases, converts the cliff into a slope the team can actually flatten.
Where drift comes from
The sources are mundane. New page types bypass the audited templates. Seasonal campaigns are built fast, often by agencies who never saw the accessibility requirements. Content teams upload images and documents at volume with no alt text step in their tooling. Third party components, chat widgets, review modules, consent banners, update on the vendor's schedule and change their markup overnight.
Each source maps to a control, which is what makes drift manageable once it is visible. Template sources need release gates. Content sources need checks inside the publishing workflow. Vendor sources need monitoring pointed at the pages where the components run, because the vendor will not announce that their update broke your focus order.
Measure defects per release, not virtue per year
The metric that changes behavior is a rate: new accessibility defects introduced per release or per hundred pages published, trended over time. A rate can be owned by the teams shipping the work. An annual score belongs to everyone and therefore to no one, and it arrives too late to connect any defect to the change that caused it.
Automated scanning supplies most of the data if it is run on a schedule against a stable page set and the results are stored, not just viewed. The trend is the product. A single scan is a list of problems; twelve scans are a story about whether the program is working, and they price the drift in defects per month, which leadership can weigh against release velocity honestly.
Sample manually where automation is blind
Automated tools catch a minority of issues and none of the judgment calls: focus order sanity, meaningful link text, error message quality, whether the alt text says anything. A small rotating manual sample, a few key journeys per month checked by someone competent, covers the blind spot at sustainable cost and calibrates the automated numbers.
Rotate the sample through the site's risk areas: new templates first, high traffic journeys always, and any page type touched by a recent vendor update. The manual findings also keep the program honest about severity. Drift measured only in automated issue counts drifts toward trivia, because trivia is what the tools count best.
Set a drift budget and enforce it kindly
Mature programs borrow the error budget idea: a small tolerated rate of new defects per period, with a clear consequence when it is spent, usually a remediation sprint before the next feature release on the offending surface. The budget reframes accessibility from perfection demanded and quietly ignored to a managed allowance that teams can plan around.
Report the budget beside the trend. When drift runs under budget for quarters, the program has evidence it works. When a team burns the budget in a month, the conversation is specific, recent, and fixable: these three releases, these defects, this sprint. That is a conversation an annual audit score can never start.
Related reading
These earlier guides go deeper on connected questions: Testing With Disabled Users: Building a Panel That Produces Evidence, Not Theater, Your Accessibility Statement Is a Metric Too: What a Credible One Contains, and Accessibility Vendor Scorecards: Scoring Third Party Tools Before They Cost You.
Frequently asked questions
What is conformance drift?
The gradual accumulation of accessibility defects between formal audits, caused by new templates, campaign pages, content publishing, and third party updates.
How often should automated scans run?
On a fixed schedule, weekly or monthly, against a stable set of key pages, with results stored so the trend is visible. The schedule matters more than the tool.
What metric best captures drift?
New defects introduced per release or per hundred pages published, trended over time. A rate can be owned and managed; an annual score cannot.
Can automated monitoring replace audits?
No. Automation misses judgment based issues like focus order and alt text quality. Pair scheduled scans with a small rotating manual sample of key journeys.