Methodology

Every number on this site is derived from two public NHTSA datasets. Here is precisely what we do to them.

Where the data comes from

We use the Office of Defects Investigation flat files published at nhtsa.gov: the complaints file (FLAT_CMPL) and the recall files (FLAT_RCL). Together they currently give us 1,569,818 complaints and 21,880 recall campaigns across 72,880 model years. NHTSA refreshes those files on a rolling basis; ours were built from the September 2026 release.

One complaint, not one row

NHTSA writes a separate row for each component named in a complaint, so a single owner report about an engine and an electrical fault appears twice. Counting rows inflates every total. We collapse rows by NHTSA's own complaint identifier (ODINO) and count each owner report once, which is why our figures match the public NHTSA lookup exactly. Component breakdowns still use every row, because that is what the breakdown is for.

Values we discard

The injury and fatality fields are free-form, and owners routinely type something else into them — a speed, an odometer reading, the number of times they visited a mechanic. We found complaints claiming 95 deaths where the text describes a transmission problem, and 99 appears to be used as an "unknown" placeholder.

A single passenger-vehicle incident cannot credibly involve more than about ten casualties, so we treat any per-complaint injury or death count above ten as a data-entry error and exclude it from totals. That affected 96 records out of 1,569,818. We never edit the complaint text itself.

Merging spellings

Manufacturers file the same vehicle under inconsistent names — "F-150", "F-150 SuperCrew" and "F 150" are one truck. We normalise cab and body suffixes, merge model names that resolve to the same identifier, and keep the most frequently used spelling for display.

How the severity label is set

The badge on each vehicle follows fixed rules, in this order:

• Do not drive — a recall on the vehicle carries an official do-not-drive advisory.
• Fatalities reported — at least one death appears in the complaint record.
• High concern — 20 or more fires, or 50 or more crashes.
• Above average — 200 or more complaints for a single model year.
• Routine — none of the above.

What these numbers are not

Complaints are unverified. Anyone can file one, nobody checks it, and manufacturers dispute many of them. High counts also track sales volume and vehicle age: a common car that has been on the road for fifteen years will out-complain a rare new one regardless of quality. Treat the severity columns and repeated failures in the same component as the real signal, and raw totals as context.

We are not affiliated with NHTSA. For a decision about a specific car, confirm against nhtsa.gov/recalls.

If a figure here does not match the source files, that is a defect in our pipeline and we want to know — see the corrections policy. The rules above are also summarised in our editorial standards.