How we test, and who pays for it

Why this site exists

Most gadget coverage is produced under conditions that make honesty difficult: review units supplied for free on an embargo, affiliate links that pay when you click 'buy', and a publishing schedule that requires a verdict within four days of unboxing. None of those things make a writer dishonest, but together they bend the average score upward by about a point and a half.

GadgetGeek started in 2021 as a spreadsheet of measurements that one of us kept for personal use. It became a site when it turned out that the measurements disagreed with the published reviews often enough to be interesting.

We have tested 214 devices since then. The average score we have given is 6.9, which is roughly what you would expect if scores were being used as a scale rather than as a marketing courtesy.

Where the money comes from

Reader memberships and a single display ad slot that is sold on a flat monthly rate and has no relationship to what we review. That is the entire business.

We do not run affiliate links. Not because affiliate income is inherently corrupting, but because it creates a measurable incentive to recommend the thing that is in stock and discounted this week, and we would rather not have to argue about whether we are immune to that.

We buy every review unit at retail, from ordinary shops, and we do not return them. Several of the devices in our reviews were bought twice, because the first unit had a defect and we wanted to know whether it was a sample problem or a product problem.

Our average score is 6.9. If that seems harsh, consider what it means that the industry average is closer to 8.4.

How a review is made

A device is bought and used as a daily driver for a minimum of 21 days. During that time we run the category's standard measurements: dBA at 30 cm for laptops, broadband attenuation on a head-and-torso rig for headphones, screen-on time under a fixed mixed-use load for phones, and reference-sensor comparison for wearables.

Measurements go into the review with their conditions attached, so you can disagree with the method rather than just the conclusion. Where we cannot measure something - comfort, software irritation, whether an app crashes - we say how many times it happened and over what period.

The score is written last, and it is absolute rather than price-adjusted. A $149 speaker and a $1,449 laptop are both judged on whether they do their job well. The price argument belongs in the verdict paragraph, where you can see it.

Corrections

We get things wrong. When we do, the correction goes at the bottom of the article with a date, and the original text stays visible rather than being quietly edited.

If a manufacturer ships a firmware update that fixes something we criticised, we retest and say so, and we adjust the score if the change is material. Three scores have been revised upward since 2021 and one downward.

If you think a measurement is wrong, tell us the method you used. We would rather rerun a test than defend a number.