How our review analysis works
Last updated: 18 September 2026
Review Analysis is the section on every car page that answers one question: is there anything in this dealer's reviews you should know about before you buy? This page explains where the figures come from, what they mean, and what we deliberately do not do.
1. Why we do not lead with the star rating
Almost every dealer looks good on stars. In our records, 9 in 10 UK dealers score 4.0 or higher on Google, so a 4.4 and a 4.7 tell you very little about what will happen if something goes wrong with your car. What separates dealers is what their unhappy customers say: whether the car matched the advert, whether faults were fixed, whether anyone answered the phone after the sale. So we read the reviews and count what they say, rather than averaging the stars.
2. What we read
We read the public Google reviews for each dealer and each of its locations. The panel covers the reviews posted in roughly the last two years, and it says so: the dates at the bottom of every panel are the window the figures are drawn from, and the review count is the number we read in that window, not the total Google lists. The figures therefore describe how a dealer is trading now rather than its whole history.
3. What we record about each review
For every review we record two things, using software that reads the text:
- What the reviewer said went well or badly, filed under one of four areas: Honesty (did the car match the advert, was the dealer straight with them), The car (faults, condition, preparation), After the sale (support, warranty, refunds, repairs, delivery) and Buying experience (communication, sales pressure, staff knowledge, price, wasted trips).
- What they came for: bought a car, visited but did not buy, or came in for a service or repair. About one review in six says nothing that places it in any of those, and those reviews are only counted under "All reviews".
A review counts once in an area however many complaints it makes there, and only reviews rated three stars or more count as praise. We record what reviewers report about their own transaction. We do not record speculation about other people's reviews.
4. The comparison with the UK average
For each area we count the share of a dealer's reviews that mention a complaint in it, and the share that praise something specific in it, and set each against the same share across every UK dealer we cover. The panel then says two things.
"Reviews of this dealer complain 2.7× more often than average about honesty and after-sales support" names the areas where we are at least 99% confident that the dealer's share is above the UK average — confident enough that, of a hundred dealers whose reviews are entirely typical, about three would show such a line by chance. The multiple is the dealer's rate over those areas against the UK rate for the same areas. "Reviews of this dealer praise the condition of its cars 2× more often than average" works the same way for praise, at 95% confidence, because a compliment that turns out to be ordinary harms nobody.
Where no single area clears its test, the complaint line looks at all complaints together: more often than average at the same 99% confidence, less often than average when the shortfall is clear, and otherwise in line with the UK average. Praise that is not above average simply reads in line: being praised less often than other dealers is not a complaint, and it is never shown as a mark against anyone.
The multiple is the plain ratio of the two shares. On a small number of reviews it tends to run high — the same dealer's other reviews usually show a smaller gap — which is why the line says about, and why nothing is said at all until the test is passed.
We need at least 30 reviews before we show anything from them: the comparison, or how those reviewers rated the dealer. Below that, the panel says there are too few reviews and shows nothing else, because a handful of ratings can differ from the UK average by chance alone.
5. Which reviews the panel opens on
The reviews closest to your own purchase are those from people who bought a car at the location you are looking at, so the panel opens on them whenever there are enough to compare. If there are not, it opens on people who bought from the dealer as a whole, and after that on all reviews. The tabs at the top of the panel let you switch, and the figures, dates and review count change with them.
6. The short version on the car page
The Review Analysis card near the top of a car page carries the same two lines as the panel, without the areas: About 2.7× more complaints than average and About 2× more praise than average, or in line with average where nothing cleared its test. It gives no verdict of its own and no rating of the dealer: each line is a summary of the counts behind the panel below, which names the areas.
The results are shown to readers who are logged in, when they press Show results. An account is free, and a dealer can open one to see exactly what is shown about it. A reader who is not logged in sees the same notice on every dealer: how many reviews we read and what the analysis compares, with the results left blank. It is the same whatever the results are, so it says nothing about the dealer it sits beside.
7. Reviews that disappear
We keep a record of the reviews we have seen for each dealer and check the listing again later. A review that a later check no longer returns is dropped from every count on this site the next time the figures are rebuilt, so a review that has gone from Google never keeps inflating or deflating a dealer's numbers. Reviews disappear for many reasons, including reviewers deleting their own, and we do not publish how many have gone from any dealer's listing.
8. Fake reviews
Fake reviews are common enough that UK law now bans them outright: since April 2025, writing or commissioning a fake review, and publishing reviews or figures drawn from reviews in a misleading way, are banned practices under the Digital Markets, Competition and Consumers Act 2024. We do not try to identify fake reviews on any dealer's listing, and we never label a review or a dealer as fake. Instead we publish a measure we can prove from our own records: what Google Local Guides give a dealer. Local Guides are reviewers Google recognises for reviewing many places. Wherever at least twenty of the reviews we checked at a listing were written by Local Guides, the panel shows their average beside the star rating, and the (i) beside it gives the average of everyone in the same reviews. Across the UK, Local Guides rate dealers a little lower than everyone else, by about a fifth of a star. We do not say why a listing's two averages differ, and we do not know.
9. What we never do
- We do not remove, reweight or rewrite reviews.
- We do not give dealers a score.
- We do not say why a review was removed or who removed it.
- We do not show a difference we cannot support with the reviews we hold.
- We do not verify reviews. They are people's own accounts of their experience, and we exclude anything we cannot count.
10. Our reviews policy
Carsnitch does not host reviews and does not accept them. Every figure on this site is counted from reviews that people posted publicly on Google, and the reviews stay where they are: we do not add, edit, reweight or remove any. When Google removes a review, or a reviewer deletes it, it leaves our counts at the next rebuild. No dealer can pay to change what the review analysis shows, to have a figure softened or a review left out, and no dealer is charged for appearing in it. Fake and incentivised reviews are prohibited by law and by us; because we hold no reviews of our own, our steps against them are to honour Google's removals promptly, to publish what Local Guides give each listing beside everyone else's average, and to check any figure a dealer or a reader questions. If you believe a review counted here is fake, or that a figure is wrong, use the link below.
We keep every review we have ever collected, every edit to it and every record of a removal, and we do not delete them. We re-fetch every listing at least once a month and rebuild the figures after each fetch, so a removed review leaves our counts within a month at most. Every panel shows the date its figures were last rebuilt.
11. Corrections and right of reply
Every figure on a panel is a count over reviews we hold, and we keep the records behind it. If you are the dealer, or a reader, and believe a figure is wrong, tell us. We will acknowledge within one working day and check the figure against our records within five working days. If it is wrong we will correct it the day we find out and tell you what was wrong; if it is right we will tell you what it counts and how. Every challenge is logged with the date, the figure, the version of the panel it appeared on, the records we checked and the outcome. A figure is never changed because a dealer asks, pays or advertises with us; it changes only when our records show it was wrong.