Open Lab Protocol · Reproducible DOM Benchmarks

How ADA Site Review Tests & Scores Platforms

Every scorecard on ADA Site Review is backed by repeatable Headless Chromium + open-source Deque axe-core WCAG 2.2 Level A & AA inspections paired with manual keyboard and mobile reflow verification.

1. What Our Automated & Keyboard Harness Measures

Our benchmark script (scripts/scan-benchmark.mjs) loads rendered pages in Headless Chromium and executes Deque axe-core across tags wcag2a, wcag2aa, wcag21a, wcag21aa, and wcag22aa (including WCAG 2.2 Target Size Minimum 2.5.8). Alongside axe-core, we probe:

  • Sequential Keyboard Focus: Verifying visible :focus-visible outlines, skip-to-content behavior, Escape-key modal dismissal, and absence of non-focusable div[onclick] controls.
  • 320px Mobile Reflow (WCAG 1.4.10): Testing desktop (1440px), phone (390px), and zoomed/reflow (320px) viewports for horizontal overflow.
  • PDF & Overlay Risk Signals: Detecting unmediated .pdf menu/intake links and third-party runtime overlay scripts (accessiBe, UserWay, AudioEye, EqualWeb).

2. Transparent 100-Point Deduction Model

Point deductions applied from a 100-point clean HTML baseline
Defect ClassDeductionExamples
Critical WCAG Violation-14 pts per nodeMissing image alt, unlabeled form input, empty button
Serious WCAG Violation-9 pts per nodeText contrast below 4.5:1, untitled <iframe>, empty link
Moderate / Minor Violation-5 / -2 ptsSkipped heading level (h1 to h4), missing landmark
Keyboard Focus Failure-12 pts (-5 Partial)outline: none without replacement, clickable <div>
Unmediated PDF Menu / Packet-6 ptsPrimary restaurant menu or patient packet served only as a PDF

3. Mandatory ~57% Automated Floor Disclosure

Open-source Deque axe-core is engineered for zero false positives and catches roughly 57% of total WCAG 2.0-2.2 issues automatically. Automated scanners cannot judge whether alt text accurately describes a dish photo or whether a complex multi-step flow makes intuitive sense to a blind guest using NVDA or VoiceOver. That is why every teardown combines empirical DOM metrics with human code review.

4. Cohort Ethics & Free Vendor Re-Tests

We name commercial software platforms, themes, plugins, and overlay vendors because thousands of merchants depend on their defaults. We never name-and-shame individual local restaurants, clinics, or small design studios; industry studies report aggregate rates only. When a vendor ships clean HTML fixes addressing a teardown, requesting a re-test is free.