How a Pruqo score is made
Every verdict is computed from public evidence, cached once, and served to everyone as the same fast, citable page. This is the whole rubric — the sources, the pipeline, the scale, and what the number does and doesn’t mean. There is no second, private ranking.
The evidence
Pruqo doesn’t have opinions about ingredients; it has citations. Five public sources do the heavy lifting, each contributing something the others can’t:
- EWG — ingredient hazard ratings for personal care (Skin Deep). The backbone of flags on skincare, sunscreen, and deodorant.
- NIH Dietary Supplement Label Database (DSLD) — the canonical record of what supplement labels actually declare, down to the serving.
- Open Food Facts — crowd-maintained label and ingredient data that fills gaps and cross-checks what brands claim.
- FDA recall and safety feeds — recalls, contamination findings, and warning letters. An open recall can cap a score on its own.
- Certification bodies (NSF, USP, USDA Organic) — third-party testing and verification. Certifications count in a product’s favor and appear as Verified flags.
Every flag on a verdict page links to its source. If it can’t be cited, it isn’t flagged.
The pipeline
A verdict moves through four steps, and the ordering rule at every one of them is: deterministic code before small models, small models before big ones.
- Resolve the product. A barcode, pasted link, or name is matched to one SKU — deterministic matching first, semantic search only when that misses.
- Read the label. The ingredient panel becomes a structured ingredient list. A model touches this step only for products never seen before; everything else is already on file.
- Join the evidence. Structured ingredients are matched against EWG, DSLD, Open Food Facts, recall feeds, and certification registries. This is a database join, not a model’s opinion.
- Flag and score. Each evidence match becomes a cited flag — Caution, Avoid, or Verified — and the flags roll up into the 0–100 Pruqo score with a plain-language summary.
The pipeline is cache-first: a verdict for a given product is identical for every person, so it’s computed once, stored, and served as a fast public page. The compute bill scales with new products, not with readers — which is what lets the core verdict stay free.
Four verdicts, redundantly encoded
Every verdict carries a number, an icon, and a label — never color alone — so it reads the same with color-vision deficiency and under harsh store lighting.
No flagged ingredients against current evidence, often with third-party certification on top. This is the tier a cleaner-pick swap points to.
A solid formula with a small number of ingredients worth knowing about. Each one is flagged and cited — the call is yours.
Several flagged ingredients, or an undisclosed blend that can’t be fully evaluated. Read the flags before you buy.
Serious, cited concerns — an open recall, a contamination history, or heavily flagged ingredients. A cleaner pick is almost always shown alongside.
What a score is not
- Not medical advice. Pruqo is ingredient transparency and price intelligence. Anything that touches your health — interactions, conditions, dosing — gets a flag and an “ask your pharmacist,” never a recommendation.
- Not a sponsor list. No brand can pay to be scored, to score higher, or to make a flag disappear. Sponsored placement, where it exists, is labeled and offered only among products that already pass the bar.
- Not alarmism. The tone is a pharmacist’s, not a preacher’s: “two flagged ingredients,” with sources — never “toxic” or “dangerous.”
- Not fixed forever. Evidence moves — recalls land, brands quietly reformulate — and scores move with it. A verdict page always shows the current evidence.
Rank is never for sale
The methodology on this page is the whole story. Pruqo earns a commission when you buy through a link — the commission never changes the pick, retailers are sorted by price, and the rules are written down where you can hold them against us.
Read the Trust Charter ↗