How we work

Testing methodology

Every showdown on this site follows the same route from “which two things should we compare?” to a published verdict. This page is that route, including the parts of it that are judgement rather than measurement.

Written and maintained by Viet HoangLast reviewedSeptember 3, 2026

The short version

Products are bought at retail with this site's own budget unless the page says otherwise. Testing is hands-on and lived-with, not instrumented in a laboratory. Anything taken from a manufacturer is labelled as a manufacturer claim. The score is one editor's judgement, expressed as a number — it is not the output of a formula.

How products are chosen

A comparison earns a place here when the choice between two products is genuinely hard to make from the outside. In practice that means one of these:

Two products that look identical on paper. Same claimed specs, very different experience.
A category full of near-clones. A dozen listings, three actual factories, and no way to tell from the product page which is which.
A gap between the marketing and the reality. A headline number that turns out to be measured under conditions you will never be in.
Reader demand. If enough people ask for a head-to-head, it goes on the list.

Commission rate is not one of the criteria. A lot of what gets covered here pays little or nothing, and some of it — software with no affiliate programme at all — pays nothing by definition. If a product is includedbecause of a commercial relationship, that would be a paid placement, and this site does not run them. See theeditorial standards.

How units are obtained

The default is simple: the unit is bought at the normal retail price, from a normal retailer, with the site's own money, like any other customer. That is the only way to be certain the unit is not hand-picked.

Sometimes a brand or an agency supplies a unit instead. When that happens it is stated on the article itself — not buried on this page — along with whether the unit was a loan or kept. A supplied unit is tested exactly the same way and buys nothing: no coverage guarantee, no review of the copy before publication, no influence on the score.

What is actually measured — and what isn't

This is an independent site run by one person, not a laboratory. Being precise about that distinction is the whole point of this page, so every figure that appears in an article falls into one of three buckets, and each is labelled:

Measured

Something checked directly, with the method described in the article — a stopwatch, a scale, a repeated run, a side-by-side comparison. The conditions are stated so you can judge how much weight to give it.

Observed

Qualitative, from using the product over time: fit, feel, build, how the software behaves, whether the annoying thing stays annoying after two weeks. This is opinion, and it is written as opinion.

Manufacturer claim

A number from the box, the spec sheet or the listing that has not been independently verified here. It is always attributed. Where a claim contradicts what was observed, the article says so.

The limits of that

Reviews are more useful when they are honest about what they cannot tell you. These limits apply to essentially everything on this site:

Sample size is one. A single unit cannot reveal how consistent a production run is, or how many units arrive faulty.
No calibrated instruments. There is no anechoic chamber, no colorimeter, no controlled thermal bench. Where precision matters, published third-party measurements are cited rather than invented here.
Long-term durability is limited to time in hand. A three-week test cannot tell you about year three. Where a product has been used for longer, the article says how long.
Software and firmware move. A verdict describes the product as it was on the date shown at the top of the article.

What the score means

Scores run from 0 to 10 with one decimal place. The number iseditorial judgement, not a computed index: there is no weighted spreadsheet behind it, no category sub-scores being averaged, and no formula that would produce the same figure if someone else ran it. It is a summary of the written verdict, and where the two ever disagree, the words are the review — the number is shorthand.

Scores are relative to the product's category and its price, not to all products everywhere. A 9.1 budget mouse is not "better" than an 8.4 flagship; it is doing more with less.

Score	Label	What it means
9.0 – 10	Exceptional	Best in its category on the things that matter, with no compromise I would warn a friend about.
8.0 – 8.9	Recommended	Clearly good. Any weaknesses are named in the review and are easy to live with.
7.0 – 7.9	Good, with caveats	Worth buying for the right person. There is a specific reason it is not higher, and it is stated.
6.0 – 6.9	Compromised	It works, but something real is wrong with it — price, build, software, or a missing feature.
Below 6.0	Hard to recommend	There is a better way to spend the same money, and the review says what it is.

When an article is updated

An article is revisited when any of the following happens:

The price changes enough to alter the recommendation.
A firmware or software update fixes — or introduces — something the review called out.
A successor product lands, or the reviewed product is discontinued.
A reader reports something wrong, and they are right.
Longer-term use contradicts the original impression. Failures after publication are added to the article, not quietly dropped.

Prices shown on the site are captured at the time of writing and go stale quickly. Always check the price on the retailer's page before buying — that is the one that counts.

What would change a verdict

Evidence. Specifically: a demonstrable error in something measured here, a fix that removes the reason for a criticism, a change in price that shifts the value calculation, or a pattern of reader reports that a fault is widespread rather than a bad unit.

What will not change a verdict: a brand being unhappy with it, the threat of losing access to future review units, or the affiliate commission on the alternative being higher. If a verdict is revised, the article records what changed and when, so you can see the revision rather than just the result.