Car complaints by model
Pick a car and see what its owners actually reported to federal regulators, ranked by the part of the vehicle they were reporting. The recalls NHTSA lists for that model year come with it. No sign-up, nothing stored, and every figure traceable to the same public file we read.
Complaint counts measure people who filed a report, not cars that failed. A model that sold in large numbers, or that was recalled, attracts reporting. Read this as what owners talked about, not as a reliability ranking.
How to read this
One complaint can name more than one component. An owner reporting a stall might have it filed under ENGINE and under FUEL SYSTEM, so the component numbers add up to more than the complaint total. That's how the federal file is built, not a fault in the arithmetic here. It's also the exact trap that inflated one of our own published figures by 42% before we caught it.
The recall list is by model year, not by car. Under the complaint ranking, the page lists the recall campaigns NHTSA keeps for that make, model and year, straight from its recall file at the moment you look. They cover every car of that run; whether a particular car had the work done is in the manufacturer's records, which only NHTSA's own VIN lookup reads. A safety recall is repaired free at a franchised dealer, however old the car is and whoever owns it now.
Absence isn't safety. A car with few complaints may be reliable, or may simply be rare, or may be owned by people who never filed anything. The federal complaint system is voluntary and most problems never reach it.
Ten complaints is the floor. Vehicles with fewer than ten records are left out entirely. Below that number a single unhappy owner moves the ranking, and this page is read by people deciding whether to buy something.
Where the numbers come from
These are consumer complaints filed with the National Highway Traffic Safety Administration, taken from the FLAT_CMPL bulk file rather than from the summary API, because the bulk file is the one that carries every row.
Model years 2010 to 2025. 645,169 unique complaints in that window, of which 626,070 belong to a vehicle with at least ten records and appear here, across 4,091 make, model and year combinations.
Each complaint is counted once per vehicle, deduplicated on the complaint number together with the make and model. That last part matters: NHTSA files tire complaints twice by design, once against the vehicle and once against the tire manufacturer, so deduplicating on the complaint number alone deletes one of two legitimate records.
Sources
- NHTSA datasets and APIs,
FLAT_CMPLconsumer complaints bulk file. The bulk file holds 2,238,354 rows, every complaint filed since 1995. Narrowed to model years 2010 to 2025 and counted by complaint rather than by row, that's 645,169. National Highway Traffic Safety Administration - NHTSA recalls API, read live for the make, model and year you pick. Free and open, no key.
How this was compiled
The bulk file is read once by tools/make-complaint-lookup.py, which deduplicates, counts, and writes a single static file. Your browser downloads that file once when the page loads, about 57 KB compressed, and every lookup after that happens on your own machine. Nothing you select is sent anywhere, because there's nowhere to send it to.
What we didn’t do
We didn't weight by how many of each model were sold, because NHTSA doesn't publish that and guessing it would put a made-up number underneath every result. So a common car will out-count a rare one even if it's the better car. We also didn't read the complaint narratives here; this page ranks the component the complaint was filed under, which is the regulator’s own classification rather than ours.