How we rank
Three picks, never fifty. Here’s exactly how we get there — because trust is the whole point.
1. Gather candidates
We build the universe of products in a category for your region from retailer APIs, affiliate feeds, and product databases — with AI research only as a fallback to catch what the feeds miss.
2. Resolve identity
The same product appears under many names. We dedupe across sources by GTIN/ASIN/MPN plus fuzzy brand/model matching into one canonical product.
3. Normalise every score
A 4.2/5 retailer rating and an 8/10 lab review aren’t comparable until we put them on a common 0–100 scale.
4. Weight, don’t just average
Independent lab tests count for more than aggregated stars. Big sample sizes count for more than a handful of reviews (we shrink low-N scores toward the category average). Recent reviews count for more than old ones. Reviews from your region count for more.
5. Add editorial judgement
An editor reviews the AI-drafted ranking and can override the order — but both the algorithm’s score and any override are recorded and shown. Nothing publishes without a human.
6. Pick three, show the working
Exactly three picks — Best Overall, Runner-up, Best Value — each with its aggregate score and clickable links to every source that fed it. Affiliate and ad relationships never influence the order.