Every grade on Ethica comes out of a deterministic calculation over cited evidence, each piece reviewed by a human. No AI ever decides a grade. This page describes the exact calculation — the engine’s, not a simplified version.
The overall score (0–100) is the weighted average of the four pillars. The letter follows fixed thresholds:
Each pillar groups three or four criteria. A criterion is scored on a published five-level rubric (0, 25, 50, 75, 100) — never a continuous scale, so two reviewers land on the same value. A criterion that does not apply is marked N/A and excluded from its pillar’s average. A whole pillar can be immaterial to the category (e.g. animal welfare for a software company): it is then excluded from the overall score and the remaining pillar weights are renormalized to total 100%. A material pillar with insufficient evidence is likewise excluded — never scored as zero — but it lowers confidence.
Every source is typed and classed. The class weighs on confidence: a regulator decision is not worth the same as a blog post.
| A | Regulators, courts | DGCCRF, RappelConso, European Commission, court rulings |
| B | NGOs | Amnesty, Fashion Revolution, BBFAW, Public Eye |
| C | Press | Le Monde, Reuters, Mediapart, investigations |
| D | Academic | Peer-reviewed studies |
| E | Company | CSR reports, the brand’s own public commitments |
| F | Contributed | User-submitted reports |
| G | Other | Blogs, opinion pieces — minimal weight |
The grade says what the evidence shows; confidence says how much evidence there is. Five dimensions, weighted:
Displayed as: High ≥ 75 · Medium 41–74 · Low < 41.
Evidence arrives continuously (regulator registers, press, NGO reports). Every item goes through a review queue where a human accepts, rejects or corrects it — source type, pillar, date. Publishing a brand requires at least one linked source; “Verified” status requires one regulator or academic source, or two concurring independent sources. A grade can change when new evidence arrives: that is the point.
A grade is a snapshot of the public evidence at a point in time — not a final moral verdict, not an internal audit. If something is wrong or outdated, the right of reply is open: every brand page says how to reach us, and any sourced correction is re-scored by the same calculation.