How these two datasets are built
Both datasets are counted from public Chinese sources and set beside the UK record. Neither is verified in the sense the rest of this site uses that word, so this page says exactly what was counted, what was thrown out, and where the counting can mislead you.
Owner figures carry no confidence label
Every figure elsewhere on this site carries one of five labels: UK tested, Official figure, Manufacturer figure, Estimate, or No data. Those labels say who verified a number. Nobody verified an owner report. Calling it an estimate would suggest we worked it out; calling it a manufacturer figure would suggest a maker stood behind it. Both would be false.
So an owner figure carries a count instead: how many reports it rests on, the dates they span, and a link to the platform they came from. Judge it on the sample, the way you would judge any survey. See the site-wide labelling policy for the five labels themselves.
The owner range ledger
Source. Autohome owner reviews. Each review carries structured fields a UK listing never has: distance covered, energy use, the range the owner got on a full charge, the season, the price paid, the month and city of purchase. We read the newest reviews for each car once a month and keep every one with a range figure.
What the sample is. Autohome shows only its most recent reviews to anyone reading the page, and offers no way to walk back through the archive. So the ledger begins the day we started collecting and grows forward. It is not a history, and a car's early quarters will always be thinner than its later ones.
What is thrown out. A report is not counted if its numbers are impossible: a full-charge range at or below zero or above 1,200 km, energy use at or below zero or above 60 kWh/100km, or an odometer under 100 km, because a car that has barely moved cannot report a range from experience. Of what remains, values outside the standard 1.5 x interquartile fence are dropped. Every published row shows how many reports were dropped alongside how many were kept.
When a figure is published. A car, a quarter and a season are published only at 5 reports or more. Below that there is no row, and no page can show a thin average. That is why a car you expect to see may be missing.
What is published. The median, not the mean, with the 25th and 75th percentiles beside it so the spread is visible. Where we hold the Chinese lab label for that car, the median is also shown as a percentage of it.
What it cannot tell you. These are owners of the Chinese-market car. Battery supplier, chemistry, motor, software and driver assistance hardware are often different in the car sold here, and Chinese driving and weather are not British driving and weather. Read the ledger as evidence about how far a claimed range falls in real use, not as a prediction of what your UK car will do. The UK WLTP figure shown beside it is the official homologated number and comes from the manufacturer's UK specification, never from this dataset.
The fault and recall archive
Three sources, three different weights. Recalls ordered through the State Administration for Market Regulation are official acts by China's regulator, and each carries a recall number, a build window and a vehicle count. DVSA recalls are the same thing for the UK, published as a single file covering every recall since 1992. Complaints on 12365auto are neither: that platform is a private company, and a complaint is one owner's report, not a finding. We keep the three apart on the page and say which is which.
Matching a Chinese recall to a UK car. Chinese notices name cars in Chinese, so we look for the Chinese names we already hold for each car we cover, and take the longest one that appears. A name immediately followed by a digit or a Latin letter does not count, because 海豹 and 海豹06 are two different cars. If no name appears, the record stays unattached rather than being guessed at.
Matching a UK recall to a UK car. The DVSA file records a legal entity in its make column, not a marque, and one importer can bring in two brands. So the brand has to line up first, and then the model strings have to contain one another. This is what keeps a BYD bus recall away from the BYD Seal: they share a make and nothing else.
Neither match reaches beyond the UK edition. The models table behind this site holds every market's cars, not only the UK's, and a Chinese name or a DVSA model string can happen to fit a car sold only in Australia or New Zealand. Both matching steps only ever consider a car with a published UK page, so a record with no honest UK match is left unattached rather than pinned to the nearest car from another market.
Pairing the two. A Chinese recall and a UK recall are put together only when they are about the same car and describe the same kind of defect, judged against a fixed list of ten systems: doors and handles, driver assistance, battery and charging, airbags and restraints, steering, brakes, software, suspension, fire, and wiring. If they share no system, there is no pair. Where both notices carry a date within a year of each other we call it the same fault in the same period; further apart, we still show it but say so.
A person checks every pair before it appears. Saying a Chinese recall and a UK recall are the same fault is a safety claim, so no pair reaches this site until someone has read both notices and approved it. Pairs that have not been checked are not merely hidden by the page; the database refuses to serve them.
What it cannot tell you. A recall on the Chinese car is not automatically a problem with the UK car, and the absence of a UK recall is not proof that the UK car is unaffected. Complaint counts are not fault rates: they reflect how many of a car are on the road and how willing its owners are to file, not how often it breaks. To check your own car, use the DVSA service and your VIN.
What we know about these sources
Chinese platforms are not neutral ground, and we would rather say so than have you find out elsewhere. Autohome was named by China's internet regulator in January 2026 over partial test information. 12365auto is a private company, not a government body, despite a name that reads like one. Paid reputation management is a documented industry in the Chinese car market, which is one reason we publish medians over samples rather than individual glowing or damning reviews, and why a fault has to appear in two independent sources before we describe it as a pattern rather than a signal.
We take no payment, cars, travel or hospitality from any manufacturer or dealer. Corrections to these datasets are listed on the corrections page like any other.