Methodology
Where the data comes from
Every figure is derived from the US National Highway Traffic Safety Administration's public APIs at api.nhtsa.gov: the owner complaint database and the recall campaign database. Data is refreshed daily. We store no personal information from complaint records.
The severity index
Raw complaint counts reward popularity, not danger - a best-selling sedan attracts more filings than a rare coupe. The severity index instead measures what kind of complaints a vehicle attracts, per 100 filings:
| Event in a complaint | Weight |
|---|---|
| Crash reported | 4 |
| Fire reported | 6 |
| Injury reported | 8 |
| Fatality reported | 40 |
Score = (4xcrashes + 6xfires + 8xinjuries + 40xdeaths) / complaints x 100. The weights are our editorial judgement, chosen so that a single fatality outweighs a large number of non-injury reports. They are not an NHTSA standard.
What Low, Medium and High mean
The bands are relative, not absolute. Every vehicle is compared against the rest of the fleet we track, using the lower and upper quartiles of the current dataset, recomputed on every update:
| Band | Severity index | Meaning |
|---|---|---|
| Low | below 35 | cleanest quarter of tracked vehicles |
| Medium | 35 to 96 | middle half |
| High | 96 and above | worst quarter of tracked vehicles |
An absolute cut-off would be misleading: the median tracked vehicle scores around 35-96, so a fixed threshold would label most of the fleet 'high' and tell you nothing. Because the bands are relative, roughly a quarter of vehicles will always be 'high' by construction - that is a comparison, not a safety verdict.
Limits you should know about
- Complaints are unverified consumer reports, not confirmed defects.
- Counts are not normalised by units sold or vehicle age.
- Media coverage and class actions measurably increase filing rates.
- A vehicle with few complaints may simply be rare.
We publish these caveats because the numbers are only useful if you know what they cannot tell you.