Editorial Methodology

We work from primary sources, not from other people's conclusions. A recommendation here is built on what the maker itself publishes: the specification, the support and lifecycle pages, the price list, the compliance documents it has signed. Those are read alongside the record a product has accumulated with the people who bought it. Never a roundup, never another site's summary of a page we could have read ourselves.

We do not run a hardware lab. There is no test bench here, and we hold no review units, bought or loaned. So you will not find a benchmark on this site presented as ours: where a figure comes from someone's test, we name whose test it was, and where a question can only be answered by living with a product for a month, we say so instead of implying we have. Admitting that costs us, because a page that names its limit reads as less authoritative than one that doesn't, and we would rather be worth checking than sound impressive.

What that leaves is a narrower job done properly: reading the primary record closely, comparing like with like, and being specific about who a thing suits and who it doesn't.

Every specific figure is traced to the party that published it before the piece goes out: a price, a support date, a benchmark score, a release year. Where the number does real work, the piece names that party. A figure we cannot confirm does not ship with a hedge attached; it comes out.

Software helps with the research and the first draft, the way any newsroom uses the tools available to it. It does not publish. Nothing reaches this site without a person reading it against the sources and deciding it is right, and the judgment and the byline stay with that person.

We favor a clear written verdict over a number. A score compresses away the nuance that usually decides whether a thing is right for you: who it's for, who it isn't, and what we'd buy instead. It also implies a measurement behind it, and we would rather give you the reasoning than a figure that sounds like a result. Where we do use a rating, we explain what it means and how we arrived at it.

Our tools answer questions with figures, and a figure is worth exactly as much as its source. Every one comes from the company that publishes it: the maker's own support page, its lifecycle table, its published price list, or a compliance document it has signed. Never a roundup, never another site's summary of that page, never a spec aggregator.

Where a maker publishes nothing, we say so. That standard costs us, and we keep it anyway: a tool that answers "no published commitment" looks emptier than one printing a confident date, and the confident date would usually be a guess wearing a fact's clothes. If a maker promises "at least four years" and nothing more precise, four years is the answer you get. We do not turn it into a month.

Where a maker publishes two figures that disagree, we use the specific one over the general one. A per-model end date beats a blanket promise made across a whole range, and we say which one we used.

Every tool prints the date its data was last checked and the schedule it is checked on. Those dates are load-bearing, not decoration: a standing maintenance review flags any tool that has passed its window, and the fast-moving ones are checked more often than the stable ones, because prices move faster than the arithmetic around them.

When a source changes what it says, the tool changes with it. When a claim can no longer be verified, it comes out rather than being left standing.

Articles may contain affiliate links, and any page that does carries a visible disclosure. They never affect the assessment, the verdict, or whether a product is recommended. See our Ethics Policy for the full disclosure standard.

Technology changes after launch: firmware fixes things or breaks them, prices move, support windows close, and a maker quietly stops selling the thing we recommended. When that materially changes our take, we update the piece with a dated note rather than leave a stale recommendation standing, and when a claim can no longer be verified at all, it comes out. We fix the record in the open; we do not make quiet edits.

Think there's something we should cover, or want to question a recommendation? Reach us through the Contact page.