What a finding is, what statuses mean, and how findings are documented.
What a finding is
The auditor's evaluation of one criterion on one sample. At most one finding per (sample, criterion) pair.
Statuses
| Status | Meaning |
|---|---|
| Success | The sample meets the criterion |
| Failure | The sample does not meet the criterion |
| Needs review | The result could not be confirmed and a human still has to look |
| N/A | The criterion does not apply to this sample (no matching content) |
| Derogation | An exemption applied, with a written justification |
| Not tested | The criterion has not been evaluated yet |
- "Needs review" and "Not tested" are no verdict: they do not advance progress or score as pass or fail.
- "Needs review" records a visit without a false Success: an ambiguous scanner result, or a criterion an AI pass could not verify.
- Failures and Derogations feed the report's issue list and the accessibility statement.
User impact
A Failure or Derogation can carry an impact:
- Low - minor friction, the task is still possible.
- Disturbing - the task is harder, slower or degraded.
- Blocking - the task cannot be completed.
- Unset reads as Unknown.
Impact qualifies a failure (not a status) and orders failures as actions in an annual accessibility plan. See accessibility plans.
Fix profiles
Who fixes a failure: development, design, content, project management, or QA. Default suggested per criterion; the auditor can override it, including to "none of these". Carried onto tickets pushed to Jira, GitHub, Linear, GitLab or Bitbucket.
Finding fields
- Status (required)
- User impact and fix profiles, for failures and derogations
- Problem description - what is wrong, observable, in plain language
- Solution suggestion - how to fix it, code-referenced when possible
- Expert proposal - the technical fix for whoever corrects it
- Affected elements, recorded by the scanner or the browser extension
- Evidence images
- Report flag with a short reason, picked up by report sections listing selected issues
- Ticket reference, once pushed to an issue tracker
Why both problem and solution
The statement and the report draw on these fields: the problem says what is wrong, the solution how to fix it. See methodology / finding quality.
AI fix suggestions
On a criterion marked Failure or Derogation, Suggest with AI drafts the finding and proposed solutions from the criterion's reference text and the auditor's existing text.
- Choices: Copy to Findings, Copy to Proposed Solutions, Replace both or Discard.
- Applied text is stored with date and model, so AI text is always distinguishable from the auditor's. Review it like scanner output. See methodology / finding quality.
- Needs the workspace AI add-on. Without it, the button explains the add-on and its cost; an Owner or Moderator can add it there, others are asked to contact one. See procedures / manage add ons.
Expert proposal
Below a failing criterion's problem and solution, Expert proposal holds the fix itself, for the developers, designers or editors applying it.
- Type: Code (markup, styles or scripts), Design or Content (a change made in the CMS).
- Markdown; code goes in fenced blocks with the language after the opening backticks, so the public page and tickets render it as code. Write and Preview toggle text and rendering.
Sources:
- The scanner, when an applied scan result's rule has a known fix (missing alternative text, form labels, button and link names, contrast, page language and title, and other common rules). It rewrites the real element in the audit's technology, and computes a compliant colour for contrast failures.
- Suggest with AI, in the section, from the recorded elements and the audit's technology. Needs the AI add-on; uses AI credits like the other work AI. Nothing is saved until Use this proposal.
- The auditor, by hand, on every plan.
Behaviour:
- A proposal the auditor wrote or edited is never replaced by a later scan.
- Tickets pushed to an issue tracker carry it, with the elements' selectors.
- In a report template, a looping issue slide shows it with
%%issue.expert_proposal%%, empty when the criterion has none, so it can also be the slide's display condition.
Above it, Affected elements lists what the scanner or the extension recorded for the criterion: CSS selector, the element's code, what fails, and measured values (contrast colours and ratio).
Where findings are edited
Open a sample from the audit's Samples tab. Criteria are grouped by topic; each card holds the status buttons, problem and solution text, impact, fix profiles and evidence images.
Scanner-pre-filled findings
With the scanner on the workspace, automated results:
- pre-fill problem and solution text where they map cleanly to a criterion;
- record the affected elements;
- write an expert proposal when the rule has a known fix;
- mark the entry as scanner-sourced.
Only Not tested criteria are filled. A one-sample scan is reviewed before applying; an all-samples scan is applied directly and reviewed after. The auditor can edit or override anything. See integrations and procedures / use the scanner.
Related
- methodology / finding quality - what makes a good finding.
- propagate status - reusing one verdict across samples.
- audits, samples.