How we decide

Every recommendation answers four questions: who is this for, why does it fit, what is the material limitation, and what should you do next.

We start from a use case, such as lighting a small desk or cooking for a family of eight, not from a product. A verdict applies only to that use case. The same product can suit one need and not another.

We collect evidence and label where it comes from: manufacturer specifications, published third-party reviews and articles, and our own editorial reasoning. We only say a product was tested when we tested it ourselves, and the review says so. We don't use Amazon customer reviews or star ratings.

When the evidence is incomplete, the product stays UNDER REVIEW rather than receiving a guessed verdict. A MAYBE always names the condition that decides it. A PASS points you to something better suited where we have one; an alternative existing is not enough reason for a PASS on its own.

We use software to notice products that are popular or changing, for example rising in a retailer's best-seller list. A trend tells us what to research next. It is never a verdict, a rating or proof of quality.

AI tools help us summarize evidence and draft text. Every claim in a draft must link to evidence we hold. An editor checks the claims, rewrites as needed and approves the exact text that goes live. Nothing is published automatically.

Prices and availability change. When we show a price, we show when it was checked. We don't promise savings or advertise discounts we haven't verified.

Reviews go out of date. When the evidence behind a review changes or expires, we mark it REVIEW NEEDS UPDATE until an editor has looked again.

Commercial relationships, such as affiliate commissions, are kept separate from verdicts. Which products earn a commission, and how much, is never an input to a verdict.