Fictitious businesses · reconstructions for a case study

Methodology

Every "before" and "after" site goes through the same automated audit, on mobile and desktop. The numbers on these pages are read from the JSON files it produces; nobody types them.

What is measured

  • Lighthouse: performance, accessibility, best practices and SEO; lab Core Web Vitals (LCP, CLS, TBT).
  • axe-core: WCAG 2.2 A and AA rules failed, by impact.
  • Weight and request count by type; images without dimensions or in legacy formats.
  • Trackers and cookie attributes; insecure http:// resources; presence of a CSP.
  • Technical SEO: title, description, viewport, language, hreflang, valid JSON-LD, sitemap and llms.txt.

Composite score

PartWeight
Performance (Lighthouse)25 %
Accessibility (Lighthouse and axe)25 %
Weight and requests15 %
Technical SEO15 %
Privacy and security10 %
Best practices (Lighthouse)10 %

Full formulas in docs/scoring.md in the repository.

The "before" sites

They are fictitious 2015-style reconstructions with 12 deliberately seeded defects (no network, no real trackers). An automated test checks that the audit detects all 12, on mobile and desktop, and that none remains in the "after" sites.

Reproduce it

corepack pnpm install
corepack pnpm audit:site https://example.com/ --device both --pdf

Source code