MACELYR VENTURES / REPRESENTATIVE WHITE-LABEL REPORT / V2
Accessibility regression triage
Controlled demonstration report for the fictional Northstar Studio campaign review page.
Executive summary
The baseline page produced zero findings from the project’s four selected HTML heuristic checks. A controlled website change then removed the primary button’s accessible name, removed the email field’s associated label, removed the informative image’s alt attribute, and changed a section heading from h2 to h3. The checker identified those four potential accessibility regressions: one critical, two high, and one medium under the demonstration’s triage model.
Each finding is presented with a stable identifier, affected element, source location, observed problem, prioritization rationale, reproduction steps, controlled evidence, suggested remediation, and developer acceptance guidance. The simulated remediation restored all four patterns. A repeated run of the same checker against the remediated page reported zero remaining findings, and that retest state is recorded below.
This proposed workflow could give development teams a consistent way to investigate selected accessibility regressions introduced by routine website releases: preserve the before-and-after evidence, route prioritized observations to developers, and document what the same bounded checks report after a fix. It supplements—not replaces—human accessibility testing and release-specific judgment.
Test scope
- Three local HTML files under direct project control: baseline, controlled regression, and remediated states.
- Four selected deterministic patterns: accessible button name, form-control label association, image
altattribute, and sequential heading levels. - Comparison of finding identities between versions, using the checker’s rule name as the stable identity.
- Local machine-readable evidence and a human-readable demonstration transcript.
- No network requests, customer data, third-party scans, production deployment, manual keyboard review, or assistive-technology review.
Severity and prioritization rationale
Critical is used when the demonstrated change can prevent understanding or operation of the primary action for users relying on accessible names. High indicates material loss of programmatic information for a form control or informative image. Medium indicates structural degradation that may impair navigation or comprehension but does not block the demonstrated primary action by itself. These labels support demonstration triage and require context-specific human review before use on real work.
Technical methodology
The demonstration runs in a controlled synthetic test environment so the baseline, introduced changes, and remediated state are known in advance. Automated heuristic checks are intentionally separated from human accessibility testing.
Run the checker against demo/baseline/index.html. The selected patterns produce zero findings.
Use demo/regression/index.html, where four known HTML changes create four detectable patterns.
Compare rule identities across baseline and regression outputs. A rule present only after the change is recorded as introduced.
The checker maps each selected rule to a fixed triage level based on the demonstrated interaction or information loss. This is prioritization, not a legal conclusion.
The remediated state restores button text, the label relationship, image alternative text, and sequential heading order.
Run the identical checker against demo/remediated/index.html. The committed result reports zero findings.
Current versus proposed capability
Current demonstration capabilities
- Controlled synthetic HTML testing
- Four selected automated heuristic checks
- Regression comparison and demonstration severity
- Developer-oriented findings
- Simulated remediation and retesting
- Sample report generation
Proposed future service capabilities
- Broader accessibility regression testing
- Manual keyboard testing
- NVDA/Chrome review
- Qualified specialist triage
- Recurring release-batch evaluation
- White-label reporting and managed retesting
Status: proposed only; not represented as operational.
Findings
Primary action button has no accessible name
RETEST: PASS“Request a review” primary action button.
demo/regression/index.html, line 16. Exact column data is not generated by the demonstration checker.
The button contains only an arrow inside aria-hidden="true", so the selected heuristic finds no accessible name.
Critical in this demonstration because the primary action’s purpose is not exposed by text the checker can treat as an accessible name.
- Open the regression file.
- Locate the line 16 button.
- Run
node tools/a11y-check.mjs demo/regression/index.html --json. - Observe rule
button-name.
Restore descriptive visible text such as “Send review request,” or provide an accurate accessible name when an icon-only design is genuinely required.
Checker message: “Button has no accessible name.”
PASS. Remediated line 16 restores visible text: <button type="button">Send review request</button>. The repeated controlled checker reports zero total findings.
Restore an accessible name for the request action. Acceptance: the control exposes an accurate name and the controlled
button-name finding is absent on retest.Email field is not programmatically labeled
RETEST: PASSWork email form field.
demo/regression/index.html, line 15. Exact column data is not generated by the demonstration checker.
The input keeps a placeholder but has no associated label or other accessible-name relationship recognized by the selected heuristic.
High because the demonstrated form control loses programmatic identification even though visual placeholder text remains.
- Open the regression file.
- Locate the line 15 email input.
- Run the checker against the regression file.
- Observe rule
form-label.
Restore a visible label whose for value matches the input id. Do not rely on placeholder text as the label.
Checker message: “Form control has no programmatically associated label.”
PASS. Remediated line 15 restores <label for="email">Work email</label> and the matching input. The repeated checker reports zero total findings.
Restore the visible work-email label and its
for/id relationship. Acceptance: the selected form-label finding is absent on retest.Informative image is missing alternative text
RETEST: PASSCampaign planning-board illustration.
demo/regression/index.html, line 13. Exact column data is not generated by the demonstration checker.
The img element has no alt attribute, so its text alternative is undefined in the controlled regression.
High because the demonstration treats the image as informative and its equivalent text is removed.
- Open the regression file.
- Locate the line 13 image.
- Run the checker against the regression file.
- Observe rule
image-alt.
Add concise equivalent text. If product owners determine the image is decorative, use alt="" instead; that content decision is outside this heuristic.
Checker message: “Image is missing an alt attribute.”
PASS. Remediated line 13 adds alt="Abstract blue shapes representing a campaign planning board". The repeated checker reports zero total findings.
Add an appropriate text alternative based on the image’s intended purpose. Acceptance: an explicit
alt decision is present and the selected image-alt finding is absent.Heading structure skips from level one to level three
RETEST: PASS“Request a review” section heading.
demo/regression/index.html, line 14. Exact column data is not generated by the demonstration checker.
The heading after the page h1 is an h3, creating a skipped level in the controlled document outline.
Medium because the structural sequence is degraded, but the demonstrated primary action remains present.
- Open the regression file.
- Review the
h1on line 11 andh3on line 14. - Run the checker.
- Observe rule
heading-order.
Use an h2 for the section so the heading hierarchy reflects the page structure.
Checker message: “Heading hierarchy skips a level.”
PASS. Remediated line 14 uses <h2 id="contact">Request a review</h2>. The repeated checker reports zero total findings.
Restore the section heading to
h2. Acceptance: the document uses sequential levels and the selected heading-order finding is absent.Retest status
The remediated synthetic page restores visible button text, a label association, an explicit image text alternative, and sequential heading order. The repeated controlled checker run reports zero findings. Machine-readable outputs are preserved in artifacts/evidence/.
Limitations and exclusions
- The checker is intentionally narrow and recognizes only four selected HTML patterns used by this controlled demonstration.
- It does not test color contrast, keyboard order, focus visibility, dynamic state, broader ARIA semantics, multimedia, documents, mobile assistive technology, or full WCAG success criteria.
- Line locations in this report were confirmed from the controlled source files; the checker itself does not generate line or column coordinates.
- Severity is a demonstration triage judgment and requires context-specific human review before use on real work.
- No manual keyboard test, NVDA/Chrome review, qualified specialist review, legal review, production monitoring, or customer-site testing occurred.
- The report is representative presentation material, not evidence of a completed commercial engagement.