OUR DATA · METHODOLOGY · POSITION
Open data, in plain English.
leastprocessed.com indexes UK supermarket products and scores them on processing level. Behind the score there's a stack of public open-data sources, regulatory registers, and a small set of editorial decisions documented openly. This page is the ‘show your working’ one.
Data sources
Everything is open and audit-able.
Open Food Facts (food)
Crowdsourced product database; the source of NOVA group, Nutri-Score, ingredient lists, retailer tags, packaging types, allergens, and barcodes for the food vertical. Licensed under ODbL 1.0; we attribute on every page. Source ↗
Open Beauty / Pet / Products Facts
Sister projects covering cosmetics, pet food, and household goods. Same ODbL 1.0 licence, same crowdsourced model, much sparser than the food set — we backfill via OFF taxonomy and manual editorial review where data is missing.
UK Food Standards Agency (FSA)
Authorised additive register (the food_authorisations dataset). Per-nation authorisation status (England, Scotland, Wales, Northern Ireland) for every E-number page. Updated regularly from the FSA register. Source ↗
Serper.dev (Google search)
We use Google search results, accessed via Serper, to map products to the retailers carrying them. We don't scrape retailer sites directly. Result: a which-supermarket-stocks-this view that wouldn't otherwise be possible from OFF alone.
Hand-reviewed primary literature
For ingredient health-claims, we use a 4-tier evidence rubric (Backed by trials / Some evidence / Mixed / Not enough yet) with live PubMed search URLs as citations and links to credentialed creators (RDNs, MDs, PhDs) who have spoken to each topic with primary-source citations. We never fabricate DOIs. We never assert a claim more confidently than the literature supports.
What's In Our Products
Detailed ingredient panels for Unilever brands across home + beauty — a rare instance of voluntary corporate transparency. Where it overlaps our index, we use it to upgrade ingredient lists and flag function notes. Source ↗
Position · why we built this
If one person can do this, government can do better.
UK and EU regulators talk, regularly, about wanting food, cosmetic, and cleaning labelling to be more structured, more transparent, more comparable across products. The standard pushback is that manufacturer diversity, retailer fragmentation, and e-commerce complexity make this intractable.
We disagree.
leastprocessed.com is built by one person, in evenings, on top of public open-data licences. It indexes 23,000+ UK food products, 60,000+ cosmetics, 15,000+ pet foods, 41,000+ household products — across nine UK supermarket chains — and gives every one of them a single 0–100 processing-level score, a transparent ledger, and a regulatory-status check. The whole system regenerates in under fifteen minutes when the underlying data updates.
That's what an independent person can do with public data, in 2026. Government with statutory powers and budget for full-time staff can do substantially more — across better packaging mandates, machine-readable ingredient lists, mandatory NOVA-equivalent classification, and consumer- accessible why-this-additive-is-here explanations.
We don't think people should have to install eight apps and read 24-point labels in supermarket aisles to know whether their breakfast cereal contains eight grams of added sugar per portion. The information is already in regulators' filings; the missing layer is the consumer-facing UX.
This site is an existence proof that the missing layer is buildable.
Methodology · how the score works
100 minus what's been done to it.
Every product starts at 100. Each piece of processing, each additive, each evidence-based detractor (palm oil, hydrogenated fat, microplastic packaging, PFAS treatment) costs the score points, transparently logged in the per-product ledger. Less processing = closer to 100.
For the full per-vertical breakdown — food's NOVA + additive + Nutri-Score weights; beauty's EWG-style hazard sum; pet's by-product penalty + named- meat bonus; home's surfactant + transparency layer; clothes' fibre composition + microplastic-shed model — see /methodology/ →
Re-use
Use this data. Improve it. Send corrections.
The derived dataset under data/dist/ in the
GitHub repository
carries the same ODbL 1.0 obligations as the OFF source data: attribute,
share-alike, and keep open.
Found a bug? An ingredient panel that shouldn't be there? An additive misclassified? A retailer wrongly flagged as carrying a product? Tell us — corrections turn around in days, not months. The editorial-policy page documents the corrections process.