Cosmetics Safety and Labelling Compliance
Netherlands
A document pipeline that turns safety data sheets into structured, regulation-aware records. Including the labelling rules a human reviewer would otherwise apply by hand.
- Client
- A Dutch medical-diagnostics group
- Industry
- Regulatory Compliance
- Period
- Jan 2025 — Jun 2025
Outcomes
80%
Efficiency gain reported by the client
- Source formats
- PDF, Excel, Word, photographed sheets
The situation
Cosmetics compliance documents are assembled by hand from safety data sheets, ingredient declarations, supplier attestations and the regulations themselves. A consultancy needed the assembly automated without losing the traceability that makes a document defensible.
What we built
Ingestion
Schema-driven extraction over PDF, Excel, Word and photographed sheets. Format-specific parsers were rejected early; suppliers do not agree on format and never will.
Domain rules
INCI nomenclature, the standardised forms some declarations require, and the vegan and animal-testing claims that turn on supply-chain facts held in different documents than the ingredient list. Those are reconciled rather than inferred.
Validation
Impurity detection and claims checking against a retrieval layer over the regulatory corpus, so a generated statement cites what supports it.
What mattered
Every extracted field carries its provenance into the generated document. Compliance output that cannot be traced to a source is not usable in an audit.
Outcome
The client reported an 80% efficiency improvement on the workflow this replaced.