How the record is built
Apparel Signals reads what fashion brands publish about their own products and turns it into a comparable record. This page explains where the data comes from, how it is classified, how the numbers are computed, and what they can and cannot tell you.
Public product pages, read weekly
Every week we visit the public product pages of the brands we cover, the same pages any shopper can open in a browser. From each product we record what the brand publishes: fibre composition, price, category, certifications and material claims, care instructions, and country of origin where stated.
- Source
- Public e-commerce pages.
- Cadence
- Weekly, per brand and per market. Each week's figures use the latest collection everywhere.
- Coverage
- Mostly European catalogues, mainly the euro area, with early coverage in Brazil. The United States and wider Latin America are next.
Clothing only
A product enters the record when the brand lists it as clothing and does not also file it under accessories, footwear, home or beauty. Shoes, bags and homeware follow their own composition and pricing logic, so the record compares like with like.
Every spelling, one taxonomy
Brands publish composition as free text, in several languages and hundreds of spellings. We map each spelling to a single controlled list of materials, so that “recycled polyester”, “poliéster reciclado” and “rPET” all count as the same thing.
Claims stay on the fibre. In a cotton and recycled polyester blend, the recycled claim belongs to the polyester.
Certifications and trade names. A certification is a scheme the page discloses, such as GOTS, GRS or Better Cotton. Trade names such as TENCEL and LYCRA are recorded separately, as branded materials.
Where artificial intelligence fits. It matches the published text to the taxonomy. The composition itself always comes from the page; where a brand publishes none, the field stays empty.
What the figures mean
Product coverage. This answers how many eligible clothing products contain a fibre or material group. Its denominator is the pure-apparel product records in the selected brand, market and reporting week. One blended product can count in several material groups, so coverage percentages can overlap.
Composition share. This answers what share of the included declared composition belongs to a fibre or material group. The pipeline sums each recorded, non-null composition percentage after canonical material mapping, then divides each material group total by that declared-composition total. It does not weight by garment sales, physical mass or catalogue size. The six material group rows in the normalized bar total 100%, apart from display rounding.
A blend example. In a hypothetical one-product catalogue declared as 60% cotton and 40% polyester, cotton and polyester each have 100% product coverage because both occur in the product. Composition share is 60% cotton and 40% polyester, or 60% Plant and 40% Synthetics at material group level.
The live example. The Europe cotton figure is 39.5% of declared composition in the latest weekly market row. It is read from the market fibre row's share field, so it is not a percentage of products.
Missing declarations and Others. A product with no declaration does not contribute to the composition denominator. Missing or invalid percentage rows are left out, and Others is an explicit mapped material group, not a bucket for missing data. Products are counted once by product id for coverage; composition rows are aggregated by their canonical material.
Typical values are medians. Medians resist the pull of a few outliers: a handful of luxury pieces lifts an average well above what a typical product costs. The median price across the European market is €40.
Every figure is dated. The numbers describe the most recent weekly collection. Past weeks are kept, so the record shows how the market moves.
Prices are in euros. Other currencies are converted with European Central Bank reference rates from the week the price was observed.
Every term is defined. The data definitions give each term on the market report and the brand reports with what it counts and what it leaves out.
What the data can and cannot say
- 01The record reflects declared composition as published by brands. Composition is read from the page, not from the garment.
- 02Disclosure varies. Where a brand does not publish a piece of information, the record shows it as not disclosed.
- 03The figures describe the brands and markets we observe, from fast fashion to luxury.
- 04The carbon figure in the footer assumes grid electricity, because our host is not listed in the Green Web Foundation's directory of providers running on renewable energy. We have asked them to register. Until they do, the figure is the higher reading rather than the flattering one.
If we misread a page, we fix it
If you work at a brand we cover and the record differs from what your product pages publish, write to us from the contact page. We check the record against the page as published and correct it where warranted.
Who owns it and who pays for it
Apparel Signals is a small independent company in Amsterdam, and the record it keeps is what it sells. The work is paid for by the people who buy the reports, from researchers and journalists to buyers and policy teams. We publish the method alongside the numbers so anyone can judge both.
How long it has run
The record has run since 2 Mar 2026, a reading in 31 of the 32 weeks since. It now covers 38 brands across 2 markets, and every week of it is kept, so a figure can be read against its own past.
Larissa Akemi
Chief executive of Apparel Signals.
Ten years as a data scientist in marketing and advertising.
Based in Amsterdam.
Follow Apparel Signals as it grows
We send an occasional email about major updates to Apparel Signals, and other news.