How SOURCE assesses evidence and coverage
SOURCE does not declare a product compliant. It records identity confidence, claim status, evidence level and what is still missing — then groups the gaps so a supplier can answer once.
Written by SOURCE Research. Published 18 August 2026. Last reviewed 18 August 2026. Not legal advice.
1. Product and supplier identity
Names, VAT numbers, GTINs and aliases are scored. High-confidence matches can be accepted automatically. Medium-confidence matches go to human review. Low confidence stays unresolved. Nothing is merged in doubt. See how SOURCE works.
2. Claims, not truth
A claim is a stated value for a property (for example recycled content) on a subject (a part, material or product). SOURCE never overwrites a claim. New information becomes a new version. Status is shown; “true” is not.
3. Evidence confidence
Values can carry a confidence level. Signal green is used for verified or active states, not as decoration.
| Level | Label | Meaning in SOURCE |
|---|---|---|
| L0 | Unknown | No usable statement yet. |
| L1 | Declared | A party declared a value; evidence is not attached. |
| L2 | Evidenced | A document or record supports the value. |
| L3 | Verified | A human review accepted the evidence. The model never sets this. |
| L4 | Traceable | The value can be followed through identity, document and permission. |
4. Missing information
Required properties come from a dataset the organisation attaches (in the demo: a named ESPR aluminium dataset). SOURCE subtracts what is already available or reusable, then groups remaining actions per supplier and product family — not one email per SKU.
5. What this is not
SOURCE does not currently classify legal Digital Product Passport applicability, complete a published passport, or replace market-surveillance judgement. Those gaps are stated so the DPP readiness page stays honest.