How we test

The three labels

Every guide, review and comparison on this site carries one of three labels, at the top of the page, before you read a word of the recommendation.

Tested Tested

We bought, borrowed or already owned the product and used it for its actual purpose. A page cannot carry this label unless it records when it was tested, the setup it was tested in, and at least one original photograph. Those three fields are mandatory — the site will not build without them.

Tested does not mean laboratory conditions. It means a real car, a real garage, and a real problem, described specifically enough that you can judge whether it resembles yours.

Researched Researched

We have not used the product. The assessment comes from manufacturer documentation, specifications, manuals, and the physical constraints of the problem. Where we say something works, we are reasoning from how the thing is built — not from having watched it work.

Researched pages are useful for narrowing a field and for understanding what actually matters in a category. They are not a substitute for hands-on experience and we do not present them as one.

Community-tested Community-tested

We have not used it, but there is a consistent, specific body of owner experience we consider credible — typically from marque forums and owners' clubs where people report the same failure or the same success across years. We say where the reports come from.

Why this matters enough to build it into the software

A label is only worth something if it cannot be applied casually. So these rules are enforced by a build check rather than by good intentions: the site refuses to compile if a page claims to be tested without the evidence fields, if prose on a non-tested page says "I used this" or "in my testing", if a recommendation has no stated limitation, or if it offers no alternative.

The same check rejects hard prices in the text. Prices change weekly; a number written in an article is wrong within a month and quietly misleads everyone who reads it afterwards. We link you to the merchant instead.

What every recommendation has to answer

1. What specific, annoying problem does this solve?
2. Does it work in normal use — not in ideal use?
3. Is it worth the money, the clutter and the look of it?

If a product cannot survive all three questions, it does not go on a list. That is also why several of our guides recommend doing nothing, or doing something free, before they recommend buying anything.

Scores

Reviews carry a Fix score (how well it solves the problem) and an Ugly score (how much it offends the eye). These are deliberately separate and never averaged. A device can be a 9 for usefulness and a 9 for ugliness at the same time — that combination is more or less the reason this site exists.

There is a third, and it was rendered on every review for weeks before this paragraph existed — which was the wrong way round, so here it is.

Value

Value is 1 to 10, like the others, and it answers one question: how much of what you pay for does the problem actually need? It is not “is this cheap”. A costly tool that ends a recurring problem outright can score higher than a cheap one that half-solves it, and frequently does.

Four things move it:

• Whether a free fix gets there first. If changing a setting, cleaning something or measuring before buying would have solved it, the product scores low however good it is — and the page says so before it says anything else.
• How often the problem happens. A once-in-five-years annoyance and a weekly one justify very different spending, and the same object can score differently on two pages for exactly this reason.
• What the next option down gives up. Where a cheaper class does the same job with a real drawback, Value reflects the size of that drawback rather than the gap in price.
• Whether the specification is checkable. A maker who publishes the figure that decides the purchase is selling something you can evaluate. One who publishes only the flattering number is asking you to guess, and that uncertainty is a cost.

A high Value score looks like a cheap instrument that stops you replacing a part that was never broken. A low one looks like a competent product bought for a problem that a ten-minute check would have ruled out.

Value never rests on price alone, and none of these scores are averaged together into an overall mark. Three numbers that disagree are more useful than one that hides the disagreement.

Where the money comes from

Affiliate commissions, disclosed on every page that carries a link. Commission rates play no part in what gets recommended or in what order. The full arrangement is set out in our editorial policy and affiliate disclosure.

Corrections

If something here is wrong, we would rather know. Tell us and we will fix it and say that we did. We do not change publication dates to make old pages look fresh; when a page is substantively updated, the updated date changes and the reason is recorded.

Those reasons are public. Corrections lists every substantive one — what was published, what it says now, how it came to be wrong, and whether a specification or a verdict changed. It includes the corrections that do not flatter us, because a list carrying only the tidy half would not be worth reading.