Nine weeks, not nine minutes
Every product on the list has been bought at retail, lived with for nine weeks by a real household, and scored against a fixed rubric. Here is exactly how that works.
The protocol, end to end
- Sourcing. We start from about 340 candidates a year — reader suggestions, retailer best-seller data, and whatever the testers noticed in the wild. Nothing enters from a PR pitch.
- Purchase. Every unit is bought at retail, at full price, under a name unconnected to this site. We never accept review samples, because a hand-picked unit is not the unit you will receive.
- Bench pass. Two days of measurement: noise floor in dB, battery under load, dimensions and weight on our own scale, and a teardown photo where the product allows it without destroying it.
- Household pass. Nine weeks in an ordinary home. The tester is told to use it or not use it as they please. We do not prompt.
- The week-three cut. Anything a tester has stopped reaching for by week three is out, regardless of how well it benched. This removes roughly half the field.
- Price history. Twelve months of pricing is pulled before anything is described as a deal. A discount that runs year-round is labelled as the standing price.
- The March question. Twelve weeks after the holidays we contact recipients of last year’s list. If an item is in a drawer, it does not return.
The scoring rubric
The score out of ten is not a vibe. It is a weighted sum of five components, published here so you can disagree with the weighting:
| Component | Weight | What it measures |
|---|---|---|
| Still-in-use rate | 35% | Whether the tester reached for it in weeks 6–9 |
| Build and repairability | 20% | Materials, fasteners, whether consumables are replaceable |
| Performance vs claim | 20% | Measured result against the number on the box |
| Price stability | 15% | How the current price sits against 12 months of history |
| Return friction | 10% | Window length, restocking fees, who pays return postage |
A product scoring below 9.0 does not make the published list. That is a high bar by design — the list is meant to be short.
What the bench can and cannot tell you
Measurement is good at settling arguments about noise, battery, weight and brightness. It is useless at predicting whether someone will keep using a thing. That is why the household pass carries more weight than every bench number combined, and why a product that wins on paper can still be cut in week three.
Conflicts and disclosures
Testers do not know which candidates carry an affiliate commission. The commercial list is merged in only after rankings are locked by the editor. Any tester with a personal or financial connection to a brand recuses themselves from that category for the year. See the advertiser disclosure for how the money side works.
This year’s guide
100 gifts people actually keep
1,200 candidates, nine weeks of testing, one question: would a real person still be using this in March?
Read the 2026 list