Global score
Each scoring axis has a relative weight. Only evaluated axes enter the global score: weights are renormalized so missing data never adds fake points. If every criterion on an axis is NA, the axis is excluded entirely.
Scoring axes and weights
Source of truth: scoring.config.ts in the product. Weights below are the raw values before renormalization.
| Axis | Weight | What it measures |
|---|---|---|
| Performance | 0.25 | Vitesse de chargement (PageSpeed/lab) et Core Web Vitals. |
| SEO | 0.25 | SEO on-page, donnees structurees et SEO technique. |
| Conversion / UX | 0.2 | Securite, mesure, reassurance et elements cles de conversion. |
| Technique & catalogue | 0.15 | Plateforme, indexabilite et profondeur du catalogue. |
| Pricing & marche | 0.15 | Positionnement prix vs concurrence (necessite des donnees concurrentielles). |
| Conformite legale | 0.1 | Obligations e-commerce FR/UE : mentions legales, CGV, RGPD, retractation, coordonnees. |
| Design & marque | 0.1 | Identite visuelle : coherence de la charte, logo/favicon et adequation au positionnement. |
| Accessibilite (RGAA) | 0.1 | Conformite RGAA 4.1 / WCAG 2.1 AA mesuree automatiquement (axe-core). Enjeu legal (European Accessibility Act) et SEO/UX. |
| Qualite P1 | 0.08 | Quick wins : liens internes, medias, mobile, formulaires et consentement cookies. |
Criteria (no fake points)
Each axis is a sum of criteria with max points. NA criteria are dropped from the denominator. Fail/warn criteria can generate a recommendation with a lever (conversion, SEO, performance, etc.).
- Performance — Performance mobile (50 pts)
- Performance — Performance desktop (25 pts)
- Performance — LCP mobile (Core Web Vital) (25 pts)
- SEO — Qualite on-page (titres, meta, Hn) (40 pts)
- SEO — Donnees structurees Produit (25 pts)
- SEO — Qualite editoriale (10 pts)
- SEO — SEO technique (PageSpeed) (35 pts)
- Conversion / UX — Connexion securisee (HTTPS) (20 pts)
- Conversion / UX — Suivi analytics (25 pts)
- Conversion / UX — Bandeau de reassurance (15 pts)
- Conversion / UX — Preuve sociale (avis) (20 pts)
- Conversion / UX — Bloc achat fiche produit (20 pts)
- Conversion / UX — Recherche interne (10 pts)
- Conversion / UX — Analyse de page CRO (15 pts)
- Technique & catalogue — Plateforme identifiee (20 pts)
- Technique & catalogue — Sitemap XML (40 pts)
- Technique & catalogue — Catalogue detecte (40 pts)
- Pricing & marche — Positionnement prix vs marche (100 pts)
- Conformite legale — Mentions legales & identite (30 pts)
- Conformite legale — CGV (20 pts)
- Conformite legale — Droit de retractation (20 pts)
- Conformite legale — Confidentialite (RGPD) (15 pts)
- Conformite legale — Coordonnees de contact (15 pts)
- Design & marque — Coherence de la charte (50 pts)
- Design & marque — Logo & favicon (30 pts)
- Design & marque — Adequation design / positionnement (20 pts)
- Accessibilite (RGAA) — Taux de conformite automatique (60 pts)
- Accessibilite (RGAA) — Multimedia (sous-titres, controles) (10 pts)
- Accessibilite (RGAA) — Anomalies bloquantes (30 pts)
- Qualite P1 — Liens internes (35 pts)
- Qualite P1 — Medias, mobile, formulaires, cookies (65 pts)
Consultant report axes (loss weights)
The 12-axis client report uses lossWeight to rank work by estimated revenue impact. This is a heuristic, not a contractual forecast (see revenue-impact.ts).
| Report axis | lossWeight |
|---|---|
| SEO | 0.2 |
| SEA / Ads | 0.08 |
| Vitesse / performance | 0.22 |
| Catalogue produit | 0.06 |
| Technologie / stack | 0.06 |
| Analytics / tracking | 0.07 |
| Pricing produit | 0.14 |
| UX / parcours d'achat | 0.16 |
| Legal / RGPD / CGV | 0.04 |
| Concurrents / benchmark | 0.1 |
| Reputation / avis | 0.05 |
| Logistique / livraison | 0.06 |
How recommendations are ranked
Priority comes from criterion status (fail = high, warn = medium) and the axis lever. The action plan also uses lossWeight so a weak mobile LCP outranks a missing legal page when both are broken.
Per-site benchmark
The benchmark collector compares detected components and screenshots to expected blocks per page type (home, collection, product) and sector rules. Insights are structured (title, evidence, recommendation, device), not free-form essays.
Panel benchmarks (gallery)
Aggregated /benchmarks/ pages publish only when a cohort has enough distinct stores. Until then, this methodology page is the public description of the model.
Limits
- PDP analysis is a sample, not every SKU on large catalogs.
- No real checkout / add-to-cart (read-only by design).
- PageSpeed can be unavailable; a lab fallback may be used or the axis marked NA.
- Sector taxonomy is heuristic; misclassification is possible on mixed catalogs.
- This does not replace a senior consultant with analytics access and stakeholder interviews.
Changelog
methodologyVersion 2026-08 — aligned with scoring.config.ts axis weights and report-axis-config.ts lossWeight.
FAQ
- Why can an axis be missing from the global score?
- If every criterion is NA (data absent), the axis is excluded and weights are renormalized. We do not invent points.
- Is the revenue impact a guarantee?
- No. It is a heuristic to rank work, not a forecast of your P&L.
- Do you place test orders?
- No. Checkout is out of scope on purpose.
- How often does the methodology change?
- When scoring.config weights or criteria change, this page should move with the code. Check the changelog section.