How we calculate the evidence score

What the score is not

The obvious thing to build would be a quality score derived from composition lines. We cannot build it without contradicting ourselves. Our own materials page shows, across five products, that a composition line says nothing about how good a fabric is: 17.95 € and 110 € carry the identical words “100 % cotton”. A quality score computed from those same percentages would refute our own article.

So we measure the one thing we can actually observe from outside: how much a brand discloses, and how much of it we were able to check. That is a real quantity, and it is not a stand-in for quality.

The seven components

Each brand can reach 100 points. Every component is either documented or it is not — we do not award partial credit for effort.

• Production disclosure — up to 25. A published supplier list scores 25, a stated country of production 13, nothing 0. A brand we have not yet checked also scores 0, and its directory entry says “not checked” rather than “none”.
• Linked sources — up to 20. Four points per source we cite and link, counted up to five sources.
• Our own material check — 15. Awarded when we have read at least one of the brand’s product pages ourselves and recorded composition, price and date.
• Fabric weight stated — 12. The g/m² figure, which almost every brand omits and which is the one number that would let two cotton jerseys be compared.
• Country stated per product — 10. Not per group, per garment. Three brands in our sample currently earn it; the rest leave the field empty. A note recording that nothing was stated does not count — that would award points for documenting a gap.
• Construction detail — 10. Knit or weave, yarn quantity, button material, seams. Added after Lacoste, which publishes more of this than anyone else in the sample while the index was still blind to it.
• Cost breakdown published — 8. Rare, and worth recording when it happens.

Evidence for the price

The second figure on each portrait compares the score against what its price band usually documents. The baselines are 30 points for €, 40 for €€, 50 for €€€ and 60 for €€€€, on the simple expectation that a brand asking more should be able to say more. A brand more than 15 points above or below its baseline is marked accordingly; everything in between counts as ordinary.

Three thresholds, not a decimal. A number like 62.4 would imply a precision that this many data points cannot support.

Where the score misleads

It rewards structure as much as goodwill. A company that owns its factories can name them without asking anyone. A brand buying from hundreds of contractors has a harder disclosure task — self-inflicted, but harder. The score cannot separate the two, and you should not read a high number as virtue.

It rewards being written about by us. The “linked sources” and “own material check” components measure our work as much as the brand’s. A brand we have researched for longer scores higher than an equally transparent brand we reached last week.

It says nothing about the clothes. A well-documented shirt can fall apart in the wash. We have not worn or washed any of them, and the score does not pretend otherwise.

Changing the weights

The weights are a judgement, and we would rather they be argued with than trusted. They are printed above in full so that anyone can recalculate a brand’s score by hand from what is shown on its portrait. When we change them, the change is recorded in the site’s version history along with the reason — the construction component, for example, was added because a brand disclosed something the index could not see.