Key takeaway
Read the test rule’s mapping before interpreting its pass. A completed automated check can leave a requirement or interaction state unresolved.
Begin with the claim the testing record must support
A product owner finds an accessibility scan with no failures and wants to describe the evaluated release as conforming. Before repeating that claim, identify the actual pages, component states, target standard and requirements examined. The result may concern one rule on one captured state. Expanding that observation into a whole-page or whole-process conclusion requires different evidence.
W3C’s WCAG 2.1 Understanding Conformance guidance describes success-criterion testing as combining automated testing and human evaluation. It also separates functional conformance from usability. That is a reason to preserve evaluation scope, not a claim that every manual check is reliable or every automated result is insufficient for its specific task. The responsible reviewer needs the test definition and the remaining questions.
Use the rule-to-requirement mapping
ACT Rules Format 1.1, published as a W3C Recommendation on 5 February 2026, specifies how test-rule outcomes map to accessibility requirements. A rule can have a conformance-requirement mapping while all of its passed outcomes still leave further testing necessary. A satisfying test can have a different mapping. Secondary requirements are also distinguished. Read the mapping for the particular rule instead of applying one universal meaning to pass.
The specification’s accessible-name example maps an image-button test to success criteria 1.1.1 and 4.1.2. In that example, failed outcomes mean the mapped requirement is not satisfied, while all passed or inapplicable outcomes still require further testing. A passing name check therefore does not settle every part of either criterion. The scope is the stated example, not a claim that all ACT rules have that outcome mapping.
Complete a coverage record for one release
This hypothetical product review targets WCAG 2.1 AA for release A8. It retains a captured page state and assumes a name test with the mapping just described. Its scan passes that rule for an image button. The expanded menu state has no retained interaction review. Another rule is inapplicable to its own test subject. None of these invented observations establishes overall conformance.
The completed record holds the broad conformance claim and assigns specific evidence gaps. It does not relabel all inapplicable results as failures. Nor does it count populated rows as a percentage of accessibility: these checks have different purposes and cannot be treated as equally complete parts of a whole site.
| Retained result | Actual scope | Completed owner decision |
|---|---|---|
| Name test passed | Image button in captured state; selected mapping still requires more testing | Retain the rule pass; mapped criteria remain unresolved |
| Expanded menu review absent | No retained interaction evidence for that state | Assign competent review of applicable requirements and state |
| Different rule inapplicable | Its own applicability condition does not hold for the subject | Keep inapplicability and rationale; no whole-page pass inferred |
| Release A8 conformance claim | Full target coverage not established by these records | Hold broad claim; return the scoped result and open questions |
Keep the state, method and outcome together
A reusable internal record identifies the release, page or component, state, target WCAG edition/level, rule/version, tool/version, method and observed result. Store the rule’s applicability and outcome mapping alongside the result. Link any human evaluation to the actual state and method examined. A manual pass without a retrievable scope is not a substitute for the missing scan context.
Keep notes about changed content and conditional states. A menu can be collapsed in a screenshot while the intended workflow requires it expanded; a sign-in or error state may be absent from a scan’s capture. These are proposed coverage questions for the product’s competent accessibility reviewer, not a complete testing procedure or findings about a real product.
The owner can now request targeted evidence instead of a second undifferentiated score. For A8, recover the original rule mapping, determine the remaining criterion review and examine the missing interaction state. Preserve earlier findings if a later release corrects them. A new release should not silently inherit an evaluation whose applicable content or behavior has changed.
Describe a test archive at the level it can support
An archive of automated rule outputs can be useful without being an archive of complete conformance assessments. Describe its actual subjects, methods, retained mappings and human-review coverage. W3C’s full-page conformance discussion is another boundary: an untested part cannot simply be excluded from a page-wide conformance claim.
Use Inventory for the evidence categories and Readiness for unresolved evaluation tasks. Keep session identities, customer content, restricted screenshots and security-sensitive application states out of an initial description. A fictional coverage table can explain the structure before any real records are considered. VOID’s permissioned introduction does not certify accessibility or authorize disclosure; samples, evaluation and licensing need their own decisions.