Data sources
Every record on this site is imported from a U.S. government publisher. Nothing is scraped from another site, and nothing is invented. Here is each source, what we take from it, how often we refresh it, and the terms it comes under.
Records currently held, counted at the last rebuild of this page. Coverage grows with each import and is never complete: of the 105,678 model years across 2,179 makes in our vehicle spine, 105,678 currently have at least one complaint, recall, investigation or manufacturer communication on file.
Complaints are reports submitted by owners and drivers to NHTSA. They are not verified and do not establish that a defect exists.
NHTSA ODI bulk flat files
Complaints, recalls, investigations and manufacturer communications
- Refresh
- Daily · last successful run
- What we take
- The complete published record for each dataset: complaint text and its component, crash/fire/injury/death flags and dates; recall campaign numbers, defect and remedy summaries and consumer advisories; investigation actions, types and status; and the summaries NHTSA publishes of manufacturer bulletins.
- What we hold
- 1,582,000 complaints · 30,240 recalls · 5,340 investigations · 207,030 manufacturer communications
- How we take it
- The bulk flat files NHTSA regenerates daily, pulled in five-year chunks by received date rather than as the full archive, so a daily refresh costs tens of megabytes instead of hundreds. Field counts are asserted on every row, so an upstream layout change fails the import loudly instead of writing misaligned columns.
- What we do not take
- Nothing that identifies a person. NHTSA already redacts complainant identity before publishing; we import the redacted record and store no contact details of any kind.
- Known limits
- Complaints are unverified owner reports. Recalls are listed against model years but apply to specific VINs. NHTSA publishes only summaries of manufacturer bulletins, not the documents themselves.
Recalls apply to specific vehicles, not to every vehicle of a model year. Check your VIN with NHTSA or your manufacturer's dealer to confirm whether a recall affects your vehicle.
Check a VIN on NHTSA.govTo file a report of your own, use NHTSA’s own reporting form. We cannot submit one for you.
NHTSA 5-Star Safety Ratings (NCAP)
Official U.S. government crash-test results
- Refresh
- Weekly · last successful run
- What we take
- Overall, frontal, side and rollover star ratings per tested configuration, the static stability factor, and the driver-assist availability flags NHTSA publishes alongside them.
- What we hold
- 641 rated configurations
- How we take it
- The SafetyRatings JSON API — the one source with no bulk export. To keep request volume polite we crawl from our own highest-activity model years outward rather than walking the whole catalogue.
- Known limits
- Not every vehicle is tested, ratings describe the configuration tested rather than every trim, stars are only comparable between vehicles of similar size and weight, and the programme changed methodology in 2011.
- How we present it
- Always labelled as NHTSA's rating and never mixed with, compared against, or folded into our own Vehicle Issue Index.
NHTSA vPIC
VIN decoding for the VIN tools
- Refresh
- On demand, when a visitor decodes a VIN
- What we take
- A fixed whitelist of decoded build attributes — make, model, model year, manufacturer, body class, series, trim, drive type, engine, fuel type, transmission and plant country — plus vPIC’s own error codes.
- What we store
- Never the VIN. The decode cache is keyed on a salted SHA-256 hash, the cached payload has VIN-echoing fields stripped, entries expire, and a VIN lookup never becomes a public page.
- Known limits
- vPIC decodes the vehicle a VIN describes, not the individual car: it knows nothing about mileage, history, ownership or whether a recall was performed.
How VIN lookups are handled is set out in full in our privacy policy.
FuelEconomy.gov (EPA and U.S. Department of Energy)
Fuel economy estimates and annual fuel costs
- Refresh
- Weekly · last successful run
- What we take
- The vehicles dataset: city, highway and combined MPG or MPGe, EV range, estimated annual fuel cost, and the engine, transmission and drive descriptions that identify each configuration.
- What we hold
- 50,242 configurations, of which 41,921 are linked to a model year in our vehicle spine
- Known limits
- EPA names vehicles differently from NHTSA and folds drivetrain qualifiers into its model strings. We link a configuration to a model year only on an exact normalised match, so some EPA rows stay unlinked and some vehicle pages show no fuel economy. That gap is real and we would rather show it than paper over it with a guess.
Terms, attribution and independence
Works of the U.S. federal government are not subject to copyright in the United States, and the NHTSA and EPA/DOE datasets above are published for public use. We re-publish them in a restructured form, attribute every page to the source it came from, and link back to the official record so you can check us.
RoadVitals is not affiliated with, endorsed by, or connected to NHTSA, the EPA, the Department of Energy, any other government agency, or any vehicle manufacturer. Where an official source and this site disagree, the official source is correct. Data is sourced from NHTSA and the U.S. Department of Energy / EPA and is presented for information only. We are not affiliated with any government agency or vehicle manufacturer.
What we add, and what we never add
Our own contribution is structure and history: normalising make and model names so the same vehicle is one page, joining four NHTSA datasets and one EPA dataset onto a single vehicle spine, snapshotting the aggregates daily because no upstream source keeps history, and computing the Vehicle Issue Index.
We do not add estimates. There are no invented repair costs, failure rates, sales figures, reliability scores or editorial verdicts anywhere on this site. Where a source has no data for a vehicle, the page says so instead of filling the gap. The methodology page documents the metric in full, including the biases we cannot remove.
Refresh schedule
NHTSA ODI files are re-imported daily and the EPA and NCAP sources weekly, all in UTC. After each day’s imports we recompute per-vehicle aggregates, then the Vehicle Issue Index, then write the day’s snapshot. The “last successful run” stamp on each source above is a real timestamp read from our import log, not a schedule. The ODI card covers four separately-scheduled imports, so its stamp is the oldest of the four — everything on that card is at least that fresh — and it shows nothing at all until all four have completed a run.
Something wrong?
If a record here does not match its source, email [email protected] with the URL and the campaign or ODI number. Other questions go to contact.