How we score

Where the numbers come from

Specifications are taken from the manufacturer's own specification page or the printed manual, never from a retailer listing. Retailers copy each other, and a wrong burr diameter or boiler type propagates across a dozen shops within a week of the first mistake. Where a maker publishes nothing, we go to the manual.

Blanks stay blank

If a figure isn't published anywhere official, we leave it empty. We don't estimate, and we don't infer it from the price.

This is the rule we care most about, because an invented number reads exactly like a measured one. Once a plausible figure is in a table nobody can tell it apart from a real one. Not you, not us six months later. An empty cell is honest and a guess is not, so you will see empty cells here.

What an empty cell does to the score

A blank is scored at the middle of its scale, not as a zero. If a row is worth 2.5 points and nobody published the answer, it takes 1.25.

A zero would be a claim. On a row asking whether a subscription offers single origin, zero says it does not — and we have not established that. The only thing we have established is that we could not find out. Scoring it zero would mark the product down for a gap in our research and quietly turn an absence into a fault.

Those rows are greyed in the working, and the panel prints how many points were imputed this way, so you can see exactly how much of any score rests on a real figure and how much on a shrug.

There is a second case that behaves differently. When almost nobody in a category publishes something — fewer than a quarter of them — the silence stops being a fact about one product and becomes a fact about the market. That row is struck out and its weight leaves the total altogether rather than being imputed, so a product is never marked down for failing to answer a question its whole industry ignores. The panel shows this as points that are not judgeable from specifications.

Between them these two rules mean a score can be read as: this is what the maker published, out of what could be known. Neither invents a number, and neither punishes a company for our own missing work.

What a score is made of

Every category is scored on its own criteria, weighted to total 100, and the overall score is the sum of those criteria — not a separate impression written alongside them. A grinder's consistency carries more weight than its footprint; a pod machine is judged mostly on the capsule system and what a cup costs to make, because that is what you live with long after the machine stops being new.

The weights are visible on each category page. If you disagree with them, you can see exactly which ones moved a product up the list.

How we rate each category

Every criterion, what it weighs, which published figure it reads and what we cannot measure:

• Accessories
• Bean-to-Cup Machines
• Coffee Bags
• Coffee Beans
• Coffee Pods
• Coffee Subscriptions
• Drip Coffee Makers
• Espresso Machines
• French Press
• Gift Sets
• Grinders
• Immersion Brewers
• Instant Coffee
• Kettles
• Moka & Stovetop
• Pod & Capsule Machines
• Pour Over
• Scales
• Turkish Coffee Machines
• Turkish Coffee Pots

What we have used, and what we haven't

Reviews are labelled. A spec-led review is built from published figures, owner reports and long-term reviews from people who bought the thing — useful and honest, but it is not the same as having lived with it. Where we have used a machine ourselves, the review says so and the writing changes accordingly.

We never present research as first-hand experience.

Editorial independence

• No manufacturer can pay for a review, a higher score, or a better position.
• Affiliate links may earn us a commission but never influence a score or a ranking (disclosure).
• Rankings are generated from the scores. We do not hand-place a product above one that scored higher.

Corrections

Specs change between production runs, and makers quietly revise them. If something here is wrong or out of date, tell us — corrections are the cheapest thing we do and the most useful.