How we test

Testing, in detail

Every product is bought or borrowed at retail, with our own money wherever possible. If a manufacturer lends us something, the review says so in the first line, and we send it back when we are done. We do not accept review units with pre-release firmware that will never ship to customers, and we do not agree to quote approval of any kind.

The minimum testing period is 30 days. For anything that wears out, a pair of running shoes or a vacuum filter, we keep going until there is meaningful wear to report: the shoe reviews published here represent 612km on a single midsole. Where a claim can be measured, we measure it: battery runtime on a fixed loop at 200 nits, noise cancelling in dB at the ear canal, surface temperatures with a contact probe.

The score

Scores are out of 10, not out of 5, because the extra resolution stops everything clustering at 4 stars. Five criteria are weighted per category and published on every review page. Two reviewers score independently, then reconcile differences above one point in a written note. We publish the disagreement rather than averaging it away.

A score below 7.5 means we would not spend our own money on it. That is the whole threshold.

Updates and corrections

Recommendations have a shelf life. Every review is re-checked annually, and sooner if the price moves more than 20 percent, a recall is issued or a successor ships. When a score changes we leave the old score visible with the date and the reason, because a review that silently rewrites itself is not a review.

Editorial ethics

No affiliate links, no sponsored scores, no paid placement in a list, no “best of” roundups that exist to capture search traffic. We do not run ads sold by a manufacturer against that manufacturer’s own category. Our funding is reader subscriptions and a single clearly-labelled display slot.