How the FaultScore works
Every grade on this site comes from one public file: the complaints American drivers have filed with the National Highway Traffic Safety Administration since 1995. There are 2,151,276 of them, and the file grows daily.
The one thing this data cannot do
NHTSA publishes complaints. It does not publish sales or registrations. That single gap decides everything about how a grade can honestly be built.
Without knowing how many of a car are on the road, a complaint count says as much about popularity as about quality. The Ford F-150 has been the best-selling vehicle in America for decades; it collects more complaints than a rare import purely by existing in greater numbers. Any site that ranks models against each other on complaint totals is telling you which cars are common, while implying it is telling you which cars are bad.
FaultScore never compares one model with another. A grade only ever says how one model year of a car did against the other years of that same car.
The three indices
Each is scored 0–100, where 50 is the reference point.
1. Complaint volume — 50% of the score
This model year's complaint count against the mean year of the same model. 50 means typical for this car; 100 means roughly double its own average year. Because it is a within-model comparison, sales volume largely cancels out — a 2015 F-150 and a 2018 F-150 sell in the same ballpark.
2. Severity — 30% of the score
The share of this model year's complaints in which the owner reported a crash, a fire, an injury or a death, measured against the rate across every complaint in the file, which is 9.69%. This is a proportion of the car's own complaints, so it needs no sales figure at all. A year that draws fewer complaints but far more dangerous ones is not let off.
3. Early failure — 20% of the score
About half of all complaints record the odometer reading at the time of failure. This index is the share of those that happened under 36,000 miles — inside a typical bumper-to-bumper warranty — against the national rate of 29.3%. A part wearing out at 150,000 miles is a car being old. The same part failing at 20,000 miles is a defect.
When too few complaints record a mileage, this index is left out and its weight is redistributed across the other two rather than scored as a zero the car did not earn.
From score to letter
| FaultScore | Grade | Model years |
|---|---|---|
| 0–25 | A | 783 |
| 26–40 | B | 1,698 |
| 41–60 | C | 3,869 |
| 61–80 | D | 2,517 |
| 81–100 | F | 539 |
Higher is worse: the score measures fault, not quality.
When we refuse to grade
A model year is left ungraded, shown as a dash, when any of these is true:
- It has fewer than 10 complaints on file.
- Its model has fewer than three years with enough data to compare against.
- It is older than the 1993 model year.
That last rule is worth explaining, because it costs us pages. The complaint file begins in 1995. Defects surface early — most complaints about a car reach NHTSA within about five years of its launch — so for a 1989 model year the record opens six years after the failures it would have captured, and only catches whatever was still on the road.
Grading those years anyway produces exactly the wrong answer. A 1989 pickup shows nineteen complaints where the 1997 shows three thousand, and a naive comparison crowns 1989 the cleanest year of the run. All that number actually records is that nobody was collecting yet. 1993 is the first model year whose first five years are at least half covered by the file, so that is the cutoff. Older model years still get a page with their real counts; they just do not get a grade the data cannot support.
A grade built on four complaints would be noise wearing the costume of a measurement. Leaving it blank is the honest answer.
What a complaint is, and is not
A complaint is an allegation made by one vehicle owner. NHTSA publishes them without verifying them. Some describe genuine defects that later became recalls; some are misdiagnoses; some are one-off failures on an otherwise sound car.
People also complain unevenly. Owners of a car with a famous problem are far more likely to file than owners of a car with a quiet one, and a story in the press can lift a model's complaint count without anything mechanical changing. That is why severity and mileage matter here alongside raw volume — they are harder to inflate.
Where the images come from
Every photograph on this site was taken by NHTSA during its own New Car Assessment Program crash tests of that exact model year, and is public domain. There are no stock photos and no manufacturer press shots. A picture of "a car like this one" would be decoration pretending to be evidence.
How current this is
NHTSA republishes the complete complaint file every day. FaultScore reloads it and rebuilds every grade on the same schedule. The most recent complaint on file was made on 23 August 2026.
Not advice
Nothing here is safety, repair or purchase advice, and FaultScore is not affiliated with NHTSA or any manufacturer. If you believe your vehicle is unsafe, stop driving it and contact your dealer or NHTSA directly.