Skip to main content
VertaaUXVertaaUX

Methodology

Audit Methodology How VertaaUX Scores Your Site

VertaaUX combines deterministic rule engines, heuristic models grounded in WCAG 2.2 and Nielsen's usability heuristics, and AI-assisted analysis across 7 scoring categories to produce actionable audit reports. This page explains exactly how. See real audit examples to understand the depth of our analysis.

The Audit Pipeline

Every VertaaUX audit follows the same six-step pipeline from URL submission to interactive report.

  1. 1

    URL Submission

    User submits a URL. No browser extension or code snippet required. The engine accepts any publicly accessible web address and queues it for immediate processing.

  2. 2

    Headless Rendering

    A Playwright-based headless browser loads the page at 1920×1080, executes JavaScript, and waits for network idle. Timeout: 30 seconds. This ensures audits reflect what real users see after dynamic content loads.

  3. 3

    DOM Extraction

    Full DOM snapshot, computed styles, ARIA tree, and viewport screenshot are captured. The ARIA tree extraction enables accessibility analysis without relying solely on source HTML.

  4. 4

    Parallel Audit Modules

    Seven audit modules run simultaneously: Usability, Clarity, Information Architecture, Accessibility, Conversion, Semantic HTML, and Keyboard Navigation. Parallel execution keeps total audit time under 30 seconds regardless of page complexity.

  5. 5

    AI Synthesis

    An LLM layer reviews borderline findings, adds context-dependent analysis, and generates human-readable explanations. The LLM does not replace deterministic checks — it supplements them for nuanced, qualitative findings.

  6. 6

    Report Generation

    Scores, issues (with severity, WCAG reference, CSS selector, confidence, and fix guidance), and an executive summary are assembled into an interactive report. Each finding links directly to the relevant standard.

Seven Scoring Categories

Each audit evaluates seven dimensions. Every category is grounded in established standards and research. Weights reflect the relative impact each dimension has on overall user experience quality.

Usability Score

20%

Standards

Nielsen’s 10 Heuristics, Core Web Vitals

Checks

Load time, touch target size (44×44px per WCAG 2.5.8), viewport configuration, font sizes, interactive element spacing

Baymard Institute

Clarity Index

15%

Standards

AI Clarity Model, Reading Ease Metrics

Checks

Reading ease, message hierarchy, brand clarity, ambiguity detection, CTA action strength

Nielsen Norman Group

IA Structure

10%

Standards

Information Architecture Patterns

Checks

Link clusters, heading hierarchy, click depth, navigation redundancy, breadcrumb clarity

NNG IA Research

Accessibility

20%

Standards

WCAG 2.2 Level AA

Checks

Color contrast (WCAG 1.4.3), alt text (1.1.1), form labels (3.3.2), keyboard navigation via behavioral probing (2.1.1), ARIA landmarks, focus management with programmatic focus/blur detection (2.4.7)

W3C WAI

Conversion

20%

Standards

Persuasive Design, CTA Quality

Checks

CTA placement and wording, page dead-ends, trust indicators, social proof, friction points

Baymard Institute

Semantic HTML

8%

Standards

HTML Living Standard, Schema.org

Checks

Section-scoped heading hierarchy, landmark roles, valid ARIA attributes, semantic element usage, structured data presence

W3C HTML Spec

Keyboard Navigation

7%

Standards

WCAG 2.1.1 Keyboard, 2.4.7 Focus Visible

Checks

Tab order, focus indicator visibility via focus/blur style diffing, keyboard traps, modal Escape key behavioral testing, skip links, roving tabindex

W3C WAI-ARIA APG

How Scores Are Calculated

Each dimension starts at 100 and deducts points based on issue severity: critical issues deduct 15 points, warnings deduct 8 points, and informational findings deduct 3 points. The composite score is the weighted average across all seven dimensions.

Deduction Scale

Critical

−15 points

Blocking issues that significantly harm usability, accessibility compliance, or conversion.

Warning

−8 points

Moderate issues that degrade experience or create unnecessary friction.

Info

−3 points

Low-impact findings and optimisation opportunities for incremental improvement.

What VertaaUX does not measure

Honest transparency: these dimensions are outside our current detection scope.

  • Brand alignment — We cannot evaluate whether your visual design matches brand guidelines or target audience expectations.
  • Content quality — We detect clarity and readability issues, but cannot judge whether your content is factually correct, persuasive, or appropriate for your audience.
  • Business-specific conversion goals — We measure general CTA quality and friction points, but cannot know your specific revenue funnel or conversion targets.
  • Subjective visual appeal — Aesthetic quality is inherently subjective. We focus on measurable usability and accessibility, not design taste.

Frequently Asked Questions

Direct answers to common questions about how VertaaUX audits are conducted and validated.

What standards does VertaaUX use?

VertaaUX grounds every audit in WCAG 2.2 Level AA for accessibility, Nielsen’s 10 usability heuristics for interaction quality, and Google’s Core Web Vitals for performance. The WebAIM Million report found that 96.3% of home pages had detectable WCAG failures — VertaaUX checks the same criteria automatically on demand.

Evidence

  • •96.3% of home pages had detectable WCAG failures in the 2024 WebAIM Million study — WebAIM Million 2024
WCAG 2.2 coverage

How does VertaaUX handle false positives?

VertaaUX separates deterministic rule checks — which produce binary pass/fail results with no ambiguity — from LLM-assisted findings, which carry explicit confidence scores. Findings below 70% confidence are labelled as informational and never promoted to critical severity. Deque research estimates automated tools detect 30–50% of real-world issues, so VertaaUX combines automated rules with AI synthesis to maximise coverage while minimising noise.

Evidence

  • •Automated tools detect 30–50% of accessibility issues without human review — Deque Systems
Accuracy & confidence scoring

How often is the methodology updated?

The VertaaUX ruleset is reviewed quarterly. Updates include new WCAG criteria as they reach Candidate Recommendation status, revised Nielsen heuristic weightings based on benchmark validation, and corrections informed by continuous precision and recall tracking. The current ruleset version and changelog are published on the benchmarks page.

Benchmark results

Validation & Transparency

VertaaUX publishes its accuracy metrics, benchmark results, and sample outputs so you can evaluate methodology claims independently.

Accuracy & Confidence

How findings are validated, confidence scored, and false positives minimised.

Benchmark Results

Precision, recall, and F1 scores across audit categories.

Audit Examples

See 5 complete VertaaUX audit reports of well-known public sites with real scores, issues, and fix guidance.

Apply this methodology to your industry

Industry-specific signals, conversion benchmarks, and audit checklists built on top of the same core methodology.