The scale
Every product we cover is scored on a 1–10 scale. A score is the weighted average of the category rubric (below), rounded to one decimal. Anything under 6.0 we don't recommend; we publish the teardown but we name the alternatives we'd buy instead.
Scores are relative to the category, not across categories: a 9.1 mechanical keyboard and a 9.1 cast-iron pan are both best-in-class for their peer group, not directly comparable.
The rubric
Weights vary by niche. For every category we publish the exact weights on the review page so you can see how we got to the number.
Tech, Finance tools, Gardening tools
Core performance (40%) · Reliability over time (20%) · Value for money (15%) · Build quality (10%) · Support + warranty (10%) · Ethics of the maker (5%)
Cooking
Cook quality (35%) · Everyday usability (20%) · Durability (15%) · Cleaning + care (10%) · Value for money (15%) · Sourcing + provenance (5%)
Lifestyle, Pets
Fit for purpose (35%) · Comfort + wearability (20%) · Durability (15%) · Value for money (15%) · Materials + sustainability (10%) · After-sales (5%)
Testing protocol
We buy the product outright where possible. Where a manufacturer supplies a unit, we disclose it inline and cross-check the unit against a second retail-bought unit before publishing. Review units we keep are disclosed; review units we return are disclosed.
Minimum testing window: 30 days of real-use testing before a score is assigned. For categories with seasonal or long-tail failure modes (outdoor gear, cookware, appliances) we extend to 90 days and re-score at 6 and 12 months.
We test against at least three named competitors in the same peer group. We publish the loss cases: the specific tests where our pick came second.
Who scores
Every score is assigned by one primary reviewer, reviewed by a second reviewer with category experience, and signed off by the editor. Disagreements between the two reviewers are published inline in the review itself — you see the dissent, not a sanitised average.
Contributor bylines list prior experience with the category (years, brands used, professional context). We don't hide who scored what.
Revisit cadence
Every recommendation is re-scored at least once a year. If the market shifts earlier (a better product ships, or a maker changes quality), we re-score immediately and the review carries a dated update log.
When a score changes, prior buyers should know. If you subscribed to the newsletter you receive a notice; the page itself carries a dated changelog at the top.
What we won't do
We do not publish sponsored scores. We do not accept payment to move a ranking. We do not hide affiliate relationships — every product page discloses the affiliate status of every outbound link.
We do not use AI to assign scores. AI assists with drafting context, pulling specifications, and cross-checking claims — scoring is a human judgement call, signed by a named editor.
Scoring a specific product
A worked example lives on every review page: the weights, the per-criterion sub-score, and the rationale for each. If a review doesn't show its working, it's not finished.
Challenge a score
If you believe we got one wrong, write to [email protected] with the specific criterion and the evidence. We publish all substantive challenges and our response, dated, on the review page itself.