Most comparison sites tell you what they picked. This page tells you how, in enough detail that you can argue with it.
How the product is working out for people who already own it, built from ratings, written feedback, sales volume and price relative to the category. It decides the order products appear in. It cannot tell you whether a claim is true.
What we found when we measured the product against its category graded topics. It is a weighted composite: the one or two topics that most cleanly separate a verified product from a claims-only one count double.
What the product exposes you to in ordinary use, at the dose or specification actually present in the box.
What it costs to make, run and dispose of, including replacement parts you will buy over its lifetime.
Whether anyone outside the manufacturer has checked the claims, and what happens to the score when nobody has.
Each category gets eight to twelve graded topics written for that product type and locked before scoring starts. It is much harder to bend a standard you wrote before you knew who would win.
Specifications come from manufacturer documentation or primary retailer listings, certifications are checked against the issuing registry, and recalls and enforcement notices are searched for every candidate.
Confirmed by an independent source, plausible but unverified, or demonstrably problematic. A missing record makes a product unverified, never defective.
When a figure is not published we estimate it, label the estimate on the page and name the baseline used, whether that is peer comparison, a category benchmark or a hard engineering limit.
Before publishing we re-read the shortlist looking for brand familiarity creeping into scores, and re-check any ranking where the top two are separated by less than three points.
Every product carries a confidence level based on how much had to be estimated and whether the gaps fall on high-priority topics. Where confidence is moderate or low we say so at the top of the score section and name exactly what is unknown.