How we score model years
Every number on CarFlawIndex comes from public NHTSA data. This page explains how we count complaints, rate model years, compute the CarFlaw Index and where the data falls short.
Last updated: November 11, 2026
Data we use
- Owner complaints (NHTSA Office of Defects Investigation), updated daily. A complaint is counted once even when it names several systems.
- Safety recalls (Part 573 defect and noncompliance reports), updated daily.
- Recall completion reports: the quarterly reports manufacturers file with NHTSA on how many affected vehicles have been repaired.
- Defect investigations and NCAP crash-test ratings from NHTSA.
We merge spellings of the same make or model (for example “F150” and “F-150”) and leave out model years that appear before a model was actually sold, which are almost always data-entry errors.
The CarFlaw verdict for each model year
We compare each model year with the other years of the same model, so a popular car is not penalized just for selling more. For every year we compute a weighted complaint score:
score = (complaints + 2 × reports involving a crash, fire or injury) ÷ years on the road (1 to 10)
Complaints involving a crash, fire or injury count three times in total. Dividing by years on the road (capped at 10) keeps newer years from looking better simply because owners have had less time to report problems.
We then divide each year's score by the median score of that model's other years and assign a verdict:
| Score vs. model median | Verdict |
|---|---|
| 2.5× or more | Avoid |
| 1.4× to 2.5× | Caution |
| 0.5× to 1.4× | Solid |
| 0.5× or less | Best |
A year is not rated when it has fewer than 25 complaints, when it is one of the two newest model years, or when the model has fewer than three comparable years.
The CarFlaw Index (0–100)
CarFlaw Index = 100 ÷ (1 + score ÷ model median)
A year exactly at the model's median scores 50. Higher is better: half the median scores about 67, twice the median about 33.
Recall repair rates
The share of affected vehicles repaired is vehicles remedied ÷ vehicles involved from the manufacturer's most recent quarterly report to NHTSA. New recalls naturally start near zero, so the “lowest repair rate” we highlight only considers recalls at least one year old.
Mileage at failure and repair costs
Mileage comes from the odometer reading owners enter in their complaint. Many recent complaints leave it blank, so we show how many complaints the chart is based on.
Repair costs are dollar amounts owners wrote in their complaints when describing a repair or quote. We only use amounts near words such as repair, replace, quote or dealer, ignore amounts that look like trade-in values, towing, rentals or payments, and show a system only when at least 8 owners mentioned a cost. They are reported experiences, not price estimates.
Limits of the data
- Complaints are submitted by the public and are not verified by NHTSA. Some problems are reported much more than others.
- We do not yet adjust for how many vehicles were sold, so comparisons are within a model, not between models.
- Recall information applies to model years; not every vehicle in a year is included. Always confirm open recalls for your vehicle by VIN at nhtsa.gov/recalls.
- CarFlawIndex is independent and not affiliated with NHTSA or any manufacturer. Verdicts are statistical summaries, not mechanical inspections or buying advice.
Questions or corrections: contact us.