How we rate gadgets
Version 1.1 deliberately separates hands-on lab-test scores from research-based editorial scores. This is a transparency policy revision, not a silent rewrite.
Two score types, clearly separated
Hands-on lab-test score
Used only after our editorial team physically tests the product and records measurements or observations under a documented test setup. The article names what was tested, how it was tested and the limits of that work.
Research-based editorial score
Editorial judgment from research — not a lab-test result. It is based on published review consensus, specification analysis and third-party hands-on reports.
A research score follows the product’s relative roundup ranking plus the cited pros, cons, verdict and comparisons in that article. If the article gives no honest basis for numerical differentiation, the product is marked Not scored.
How research scores are derived
- Establish the article’s ranking or buyer-role verdict from its cited evidence.
- Weigh the stated advantages, limitations, price context, availability and comparisons against nearby picks.
- Assign a whole- or half-star score without inventing decimal category sub-scores.
- Check that a lower-ranked product does not score above a higher-ranked product in the same roundup.
Article-level badges may use a 10-point display, while product blocks use a five-star display. Both summarize published editorial reasoning; neither converts research into a lab result.
The public scale
Anything at or below 2.5 stars is not recommended.
Evidence hierarchy
- Documented first-party specifications and policies
- Regulatory filings, standards documentation and retailer status
- Credible independent reporting and third-party hands-on reviews
- User-pattern signals, treated cautiously and never as laboratory proof
Face-Offs use weighted points
Before comparing products, the article publishes criteria totaling 100 points. Each round names the winner and shows the arithmetic. When physical test evidence is missing, the framework can explain a use-case decision but cannot invent a performance score.
Change policy
Scores may change after a material firmware update, a completed retest, stronger research consensus or a corrected factual basis. Any change receives an absolute date and explanation in the article’s Recent updates block and, when material, the public corrections log.
Version 1.1 · September 22, 2026: added the explicit two-score policy and research-score derivation rules. Version 1.0 · September 19, 2026: introduced the original unified rubric.