Data and licences

Emollient places a product against a measured population rather than grading it. That population is a derived database of Open Beauty Facts, and the Open Database License asks for the derivative to be published under the same terms. It is published here.

The derived database

Licensed under the Open Database License v1.0. Contains information from Open Beauty Facts, which is made available under the ODbL.

• reference_distribution.json. The distribution every percentile sentence in the app is computed from: 525 held-out labels, counts of labelled fragrance allergens and EU restricted substances, the share carrying an undisclosed mixture, and the median list length.
• holdout_labels.json. The held-out label texts themselves. Product codes ending 8 or 9, which contributed nothing to the matcher's vocabulary, so the measurement was never run against its own training data.
• frequency.json. How often each ingredient name appears across the analysed labels.
• vocabulary.json. The ingredient name vocabulary, with the frequency rank that decides what gets researched first.

Computed 23 September 2026.

Every other source, and its position

• EUR-Lex and the Publications Office. Regulation (EC) No 1223/2009 Annexes II to VI, Commission Decision (EU) 2019/701, REACH Annexes XIV and XVII. Reused under Commission Decision 2011/833/EU on the reuse of Commission documents. The consolidated text is a documentation tool with no legal effect; the authentic text is the Official Journal version, which the app cites alongside every line.
• IARC. The classification, the volume and the year are facts and the app prints them. IARC permits reproduction of extracts for non-commercial purposes, and a paid app is not that, so the verbatim monograph sentences are withheld until written permission is granted and the app links the monograph instead.
• ECHA. The SVHC candidate list, reproduced with the source acknowledged per entry, as ECHA's terms require. ECHA's own site is behind a bot challenge, so entries were read from Internet Archive crawls of ECHA's pages and each one carries its own snapshot date.
• Health Canada. The app carries the substance name, the status and the instrument title, which are facts. No verbatim Hotlist text is reproduced, because commercial reproduction of Government of Canada content needs written permission.
• EUCOSMETICS inventory. CC BY 4.0, DOI 10.5281/zenodo.3959386. It derives from Commission Decision 96/335/EC, repealed with effect from 8 May 2020. The app says so on every function it prints from it, because it is the only open source of ingredient function and it is not current law.
• SCCS. Opinion titles verbatim, with the opinion reference and the date. A conclusion lives in the body of a PDF, and a paraphrase of one would put a safety verdict in this project's voice and then attribute it to the committee, so no conclusion is rendered as a quotation unless it carries the committee's own words. SCCS opinions are Commission documents, reused under Decision 2011/833/EU.

Sources that are closed, and are not used

EWG Skin Deep, INCIDecoder and the PCPC INCI dictionary are not used, quoted or derived from. Where a jurisdiction blocks automated access and offers no open dataset, the app records the absence rather than approximating the content.