How we score
The TechWorth score is a value verdict, not a quality rating. It answers one question — is this a good buy, at this price, in this condition, today — and it changes when the price does.
What the score is
A score out of 100 for one specific version of a device: this storage size, this condition, at this price, on this date. Change any of those and the score changes. The same phone at a good price and at a bad price are not the same offer, and a single score attached to the model name would be meaningless.
What it is not
It is not a review, and it is not a rating. Nobody has scored the camera. It is not an average of anyone’s opinion, and it is not a measure of how good the product is — a superb device at a silly price scores badly, and a modest one at a keen price scores well.
For that reason we never publish it as a star rating in the code search engines read. Doing so would put our own number where a customer review belongs, which is misleading to a reader and against Google’s rules for review snippets. Where the score appears in structured data at all, it appears as what it is: a computed property, with this page linked as the method. An automated test fails our build if a rating ever appears anywhere on the site.
What goes into it
Six components, each computed from stored data rather than judged:
- Price position. Where this price sits within the prices we have actually observed for the same device in the same condition, and how far it sits from the depreciation curve we model for it.
- Remaining useful life. How long it is likely to keep receiving security updates, adjusted for battery state and whether the hardware still does the job.
- Performance adequacy. Whether it clears the practical thresholds an ordinary UK user actually needs cleared, rather than how it benchmarks.
- Value retention. What we model it being worth in a couple of years, as a proportion of what it costs now.
- Repair economics. Whether parts exist, what a repair costs as a share of the device’s value, and how repairable it is.
- Risk and warranty. The warranty, the return window, the condition grade, and what kind of seller it is.
Each component is weighted, and the breakdown is always shown alongside the total — a score you cannot take apart is a number you have to take on trust.
The weights are published and versioned
They live in the database as an immutable, dated set, not in the code. When we change them we publish a new set rather than editing the old one, so a score given last month can still be reproduced with the weights that produced it. Nothing about a score is decided by whoever last edited a file.
Confidence is reported separately, and sometimes we publish nothing
A score built on a handful of price observations over a few days is not the same as one built on a long history from several shops, even when the two numbers are identical. So the confidence is shown next to the score, with what it is based on.
Below a threshold of supporting data we do not publish a score at all. A number with nothing behind it is worse than an admission that we cannot say yet.
No shop can pay to move a score
None of the six components can see commission, and there is no component that could. The code that computes the score has no access to the commercial layer whatsoever, and a test in our build fails if that ever changes. How we make money sets out the arrangement in full.