How we count

One distinct NHTSA recall number (CAMPNO). The recalls file has a row for every campaign, product and component, so rows are never counted as campaigns. A campaign is dated by the day the manufacturer's Part 573 report was received (RCDATE), not the day it was posted.

The manufacturer's estimate of potentially affected units (POTAFF), taken once per campaign before summing. Equipment, tire and child-seat campaigns count units, not vehicles. Campaigns without a reported count show "Not reported" and are left out of sums; captions say how many.

One NHTSA complaint number (ODINO). The complaints file repeats a complaint on a row for each component it names, and before December 15, 2002 for each product, so its 2,252,683 rows hold 1,623,694 complaints. A complaint naming two components counts once in each component's total, so component shares can add to more than 100%.

One NHTSA action number (for example PE24012: a preliminary evaluation opened in 2024). An investigation "led to a recall" when NHTSA's file links it to a campaign number.

One distinct manufacturer communication (NHTSA ID) in the communications files: technical service bulletins, over-the-air notices and other messages to dealers. A bulletin is not a recall.

The company that filed the Part 573 report (MFGNAME), mapped through the table below so that one company's name changes count as one manufacturer. Vehicle pages use NHTSA's make names (Toyota and Lexus are separate makes of one manufacturer).

Manufacturer, make, model, model years, vehicle type, component and years filed are kept in the page address and, for 30 days, in a first-party cookie, so the next page opens with the same view. A page applies the ones that make sense for it and says which it does not use; the manufacturers page highlights the chosen manufacturer instead of narrowing the comparison to it.

A recall matches the make, model, model-year and vehicle-type filters when one product it covers meets all of them together. Years filed means the year of the Part 573 report for recalls and the date NHTSA received a complaint for complaints.

Grouped from the body class NHTSA's vPIC decoder gives for each vehicle's most common complaint VIN: cars (sedan, coupe, hatchback, convertible, wagon, roadster), SUVs, pickup trucks, vans and minivans, commercial trucks and buses, motorcycles, and RVs and trailers. About 99% of complaints have a body class; about half of recall product rows do, mostly the light vehicles, so a vehicle-type filter on recall pages leaves out recalls of undecoded vehicles.

For the same years and filters, owner complaints matched to a manufacturer by the manufacturer named in the complaint, set beside the vehicle recall campaigns it filed. Complaints per recall divides one by the other for manufacturers with at least 10 campaigns. It is a comparison of two counts, not a defect rate: sales, fleet age and reporting habits all move it.

NHTSA publishes only the complainant's city and state. Each city is placed at its centre from GeoNames postal places (CC BY 4.0) and the Census Bureau's 2024 gazetteers; about 97% of complaints, and 99% of the last five years', can be placed. Street tiles are OpenFreeMap, from OpenStreetMap data.

Rates are complaints per 100,000 residents: Census Bureau estimates for July 2024 for states and counties, and for a city the larger of its ZIP areas' population (ACS 2023) and its own estimate, since owners write the postal city they live in. A postal city that spans several counties shares its complaints among them by population. A ZIP area shows its postal city's rate. Rankings need a minimum population (25,000, 100,000 or 500,000), and volume is always shown beside the rate.

Fatal crashes come from NHTSA's Fatality Analysis Reporting System, every crash on a public road in which someone died within 30 days, 2015 to 2024, at its recorded coordinates; county rates are deaths per 100,000 residents a year. A crash matches the filters when one of its vehicles does; about 92% of vehicles match a make and model on file, mostly through their VIN. Automation crash reports are those filed under NHTSA's Standing General Order since 2021 by manufacturers and robotaxi operators, deduplicated by incident and placed on the reported city; they are not verified.

The complaint text filter searches every owner's own description, one document per complaint, with word endings ignored (stall finds stalled and stalling). Every word or "quoted phrase" must appear. It narrows complaints only: rising issues and the recall side of comparisons are not text-matched.