How we choose the best stuff
Do the research. Make the call. Explain the call. Here is what sits underneath each of those three sentences.
The process
- 01
Define the buyer, not the product
Every guide starts with who is asking. 'Best TV' means best for a household watching in the evening — not best measured brightness, and not best value per square inch.
- 02
Build a shortlist that a sane person would consider
We cut anything discontinued, anything without independent measurements, and anything whose review sample doesn't represent the size or trim people actually buy.
- 03
Weight the criteria before looking at results
Deciding what matters after seeing which product wins is how motivated reasoning gets into a buying guide.
- 04
Translate the specs
A number is only useful if we can say what it does to your evening. If we can't, it doesn't go in the guide.
- 05
Name one winner
A ranked list of nine is a way of avoiding the decision the reader came for. We name one, then explain who should buy something else.
- 06
State the catch
Every product has one. A guide without a catch is an advertisement.
- 07
Re-check on a schedule
Prices, stock and model years move. Each guide carries the date it was last checked, and stale guides get marked rather than quietly left up.
What the evidence badges mean
Every guide displays only the statuses that are actually true for it. If a badge is absent, we are not claiming it. There is no greyed-out “not tested” badge designed to imply that we nearly tested something.
- Researched
- We read the independent lab measurements, the manufacturer spec sheets, the teardowns and at least a year of owner complaints. This is the baseline for every guide and it is not a lesser status — it is an honest one.
- Hands-on tested
- Someone here physically used the product. The guide says for how long and under what conditions. Review loans are disclosed and returned.
- Owned
- Someone here bought it with their own money and lives with it. This is the strongest status we have, because it is the only one that surfaces month-nine problems.
- Expert-tested
- A named outside expert or laboratory tested it and we cite them directly. We do not launder someone else's testing into an implication that we did it.
- Reader-validated
- Enough owners reported the same thing that we believe the pattern. Useful for reliability, useless for picture quality.
What would change a verdict
A new model that measurably beats the winner. A price move large enough that the value calculation flips. A reliability pattern in owner reports that we can corroborate. Or — and this one is the most common — the discovery that we weighted the wrong criterion in the first place.
When that happens we do not quietly edit the page. The change is dated, the previous winner stays visible, and where it matters it gets a “We changed our mind” note explaining what we got wrong. A publication that never publicly reverses itself is either extraordinarily lucky or not looking.