How we compare and score products

This page explains how every comparison on FairlyRanked is made. The same rules apply to every category, and none of them read prices, retailers or commission.

How we choose the models

Each category covers a defined range of products, described in the introduction at the top of its page. Within that range we include every model its makers currently sell in the UK, not a shortlist, plus recently discontinued models and models that have been announced but are not yet on sale, each clearly labelled.

• Families and variants. A model family is one product in several colourways or editions. Each colourway or edition is a variant with its own model code. Scores, badges and product pages belong to the family; the compare tool compares variants.
• The standard model. Where a family’s variants differ in a figure, such as a case size or an edition’s weight, the family’s own figures, bars, drawing and score are those of its standard model: the plainest edition in the size the maker sells under the family’s name, named on the family’s row and product page, with each category’s notes saying how it is chosen. Every variant’s own figures are shown on its card and in the compare tool.
• Discontinued models stay on the site, labelled Discontinued, with the date when we know it and a link to the successor. They keep their score but never win a badge, a quick answer or a place among suggested alternatives.
• Upcoming models are labelled Upcoming. They keep their score but win nothing until they go on sale.
• Availability is stated, never assumed. “Not sold in the UK” appears only where a source shows it; where we have not checked, the page says “UK availability not checked”.

By default a category page shows the models on sale in the UK now. A checkbox above the list shows the others too.

Where the specifications come from

Every value on a page has a source: the web page we read, and the date we checked it. We use the maker’s own information first: product pages, manuals and specification sheets. Where the maker does not publish a value, we use a major retailer’s specification page or another reliable published source, and the page names it.

• A value we cannot source is left out, and the page says “No data”. We never estimate.
• The recommended retail price (RRP) comes from the maker or a major UK retailer’s list price, never a discounted price.
• Each product page lists its sources with the values each one covers; each category page lists every source it uses, with the date it was last checked.

How we check them

A person reviews every change before it is published. When a category is added, a sample of its products is also spot-checked against their sources.

Values a person has set by hand can be locked, so that no automated update can change them. How AI tools help with the research is explained on How we use AI.

How often we update

For now, we check the specifications by hand. Once our automated checks run, each category will have its own schedule, weekly, twice a week or monthly, depending on how quickly its products change.

Each category page shows when its specifications were last checked and a list of changes, such as models added and values corrected.

How the overall score works

Each model family gets an overall score from 0 to 10, worked out from its specifications alone.

1. Only attributes with a weight count. Each category gives a weight to the attributes where one direction is clearly better for a buyer. Attributes without a weight are shown, filtered and compared, but never scored. The weights for every category are in the table below.
2. Each value becomes a number from 0 to 1, where 1 is best:
   • Numbers are compared across the category’s models on the UK page: the best value gets 1, the worst 0, and the rest in proportion. Where lower is better, such as weight, the scale is reversed. When every model with a value has the same value, each of them gets 1.
   • Yes or no features get 1 for yes and 0 for no (reversed where no is better).
   • Graded values, such as an energy class, go from 0 for the worst grade to 1 for the best, evenly spaced.
   • Water protection, where a category scores it, is graded by the second digit of the IP rating, the one for water (IPX4 < IPX5 < IPX6 < IPX7 = IP67 < IP68; the first digit, for dust, does not count), and compared across the category’s models like a number.
3. A model missing a scored value is not ranked. Where the maker publishes no figure for a value that counts, the model gets no score: its page says “Not ranked” and names the missing value, it comes last when models are sorted by score, and it is never suggested as an alternative. Its other figures still count when the other models are compared.
4. An attribute that does not apply to a model is left out of its score, neither 0 nor missing. For example, features that need a phone connection do not count against a model without one.
5. The score is 10 times the weighted average, rounded half up to one decimal. A model with no scored attribute that applies has no score.

The score never reads prices, offers, retailers or commission.

Score bands

The score compares a model with the others in its category, so a model can score low because the others have more of what is scored. Beside every score a band says where the model sits in its category, in words:

• Top of the range: a score of 6.5 or more.
• Mid-range: from 3.5 to 6.4.
• Entry: below 3.5.

Bands are worked out from the score as shown, to one decimal. They describe a model’s place among the others in the category, not how good it is on its own. A model that is not ranked has no band.

Weights by category

Badges and quick answers

Badges such as “Best overall” and quick answers such as the lightest model come from the data, never typed in.

• Only models you can buy in the UK compete: at least one variant must be confirmed on sale here, and the model must be neither upcoming nor discontinued. Other models keep their score but never win.
• A model that lacks the figure an award needs is left out of that award only: a model that is not ranked is left out of Best overall and Best value, a model with no RRP for its standard model is left out of Best value, and a model missing a figure is left out of the award for that figure. It can still win the others.
• Nothing is awarded when fewer than two competing models have the figure it needs.
• Ties share the badge. Scores are compared as shown, to one decimal.
• A model shows at most two badges.
• The biggest and smallest compare the volume of the case (length × width × height). Most features counts the yes-or-no features where yes is better.
• An award for the highest or lowest value of one specification (the smallest, lightest, biggest and most powerful, and badges such as Longest battery life) considers every variant on sale in the UK with its own figures, not just the standard model, so a smaller edition of a family can win. The quick answer then names the family and that model’s size (the family’s name, then “40 mm”), and the badge goes to the family. Best value and most features use the family’s figures.
• Best value is the family’s score per pound of its standard model’s RRP, so the score and the price belong to the same model, and says “based on RRP” until we show retailers’ prices.

Suggested alternatives

Each product page can suggest up to three models in each of three lists. Only models that compete for badges and are ranked are suggested, and nothing is suggested for a model that is not ranked.

• Better specified: a score more than 1.0 point higher, the nearest step up first.
• Cheaper: a score within 1.0 point and a lower RRP, the cheapest first.
• Similar: a score within 1.0 point that is not cheaper, the closest in score and size first.

Prices only choose the Cheaper list; they never change a score.