Editorial policy

Who decides what gets tested

The desk does. Test subjects are chosen on how many people use the product and how consequential the number it reports is — a calorie estimate or a sleep stage changes behaviour, a step count mostly does not. Nobody outside the desk can put a product on the bench or keep one off it, and there is no mechanism by which paying us would achieve either, because there is no mechanism by which anyone pays us.

What goes on a page

• A figure appears with its unit, its sample size, the date it was measured and the thing that produced it. All four. Our content model refuses to accept a figure without a source, which is a deliberate obstacle and not a bug.
• The verdict is stated in the first lines, in plain words, before any table. A reader who stops after two sentences should still have the answer.
• Products are named with a link to their own official site on first mention. We link to the maker, never to a store, never to an affiliate URL, and never to a review.
• Where we recommend something, we also say what it is worse at. A product with no drawbacks has not been tested, it has been described.

What we will not publish

• A measurement we did not make. There is no house style that lets an estimate become a figure by being written confidently.
• A negative capability claim based on silence. "The app does not do X" gets published when the maker says so or we tested for it and it was absent — not because a feature page failed to mention it. An invented limitation is a factual error a competitor can disprove with one screenshot, and it is worse than saying nothing.
• A renewal or introductory price the seller has not stated. See method clause M-08.
• A vendor's own accuracy figure presented as independent validation. See clause M-09.
• Photographs of food, gear or people. This site has no photography and no stock imagery, deliberately.

Sources and labelling

Every figure carries its provenance in the same row. Ours says so. Somebody else's names them. A manufacturer's own claim is labelled as a claim by the manufacturer, every time, including when we think it is probably right.

Where an independent result has been reproduced by a second, unaffiliated lab on its own samples, we say that explicitly, because cross-lab replication is a materially different thing from a single study and readers deserve to know which one they are looking at.

Conflicts

No affiliate accounts with any company we cover. No advertising. No sponsored content. No review units, no press accounts, no embargoed briefings. Every product bought at retail. Analysts publish individual conflicts statements on their own pages and update them with dates rather than deleting the old text.

Right of reply

A company we have measured can respond, and we will publish the response alongside the figure if it is factual and signed. What it cannot do is see the page first. Where a response causes us to re-run something, the re-run result stands whichever way it goes, and the whole exchange gets an entry in the corrections log.

Software assistance

We use software for arithmetic, for drafting and for checking our own prose. We do not use it to produce a number. Every figure published here traces to a physical measurement made by a person named on the page, and an analyst signs off on the run before it goes out. A model does not sign anything.

Updating a page

Prices and availability get re-checked and the date on the page moves. A figure that changes because we re-measured gets a log entry. A figure that changes because we were wrong gets a log entry that says so. Silent edits to numbers are the one thing that would make everything else here worthless, so they do not happen.

Contact

Editorial: desk@referenceplate.com. Corrections: corrections@referenceplate.com.