Understand results

Read findings together with coverage

A finding describes one normalized accessibility concern. Coverage describes what the engines could and could not evaluate. You need both to understand the result.

Automation levels

Level	Meaning	Can contribute to detected failures?
`deterministic`	The implemented check can make a reliable machine decision for its exact scope	Yes
`partial`	The check reliably detects some failures but cannot prove the whole requirement	Yes, for detected failures only
`heuristic`	A signal prioritizes human investigation	No deterministic-failure count
`manual`	A person must assess the requirement	Never an automated pass

An automated pass for a partial check means only that its detectable failure pattern was not found. The report expresses this as no_detected_failure, not full requirement compliance.

Finding identity and occurrences

Findings are grouped by normalized rule and a canonical semantic locator. The fingerprint deliberately excludes source hash, severity, and profile mapping so a stable issue can survive normal content revisions.

Use:

finding ID for the immutable record in one scan;
fingerprint for compatible baseline comparison;
rule ID for remediation automation and reporting;
occurrence locators for affected elements, pages, objects, or document structures.

Reports include occurrence totals and at most 20 representative occurrences per finding. Do not assume the sample is the complete list of affected nodes.

Severity is prioritization, not legal impact

blocker, critical, major, minor, and info are technical triage levels. Customer rule packs can override severity without changing the underlying rule, fingerprint, or standards mapping.

Suppressions retain evidence

A suppression does not delete the finding. It attaches an actor, reason, scope, and optional expiry. This preserves auditability and allows an expired suppression to re-enter release evaluation.

Always display or export suppressed findings separately. Hiding them entirely makes it impossible to review technical debt and stale exceptions.

Coverage outcomes

Coverage items explain rule execution:

passed or failed for executed deterministic scope;
not_applicable when the rule does not apply to the observed asset;
reduced_coverage when resources or analysis were incomplete;
not_executed when a required check did not run;
manual when human review is required.

Requirement-level reports can additionally distinguish no_detected_failure, review_required, and unsupported according to mapping coverage.

Practical review order

Resolve incomplete or not-executed coverage before trusting a clean-looking count.
Fix unsuppressed deterministic and partial failures by severity and affected scope.
Investigate heuristic signals.
Complete every applicable manual checklist item.
Review active suppressions and upcoming expirations.
Re-scan and compare against a compatible baseline.

This order turns the report into remediation work rather than a static scorecard.