50 years of U.S. road-safety data, unified and ready to analyze.
About 1.87M fatal crashes from 1975–2024 and 328K+ NHTSA recall records, linked and ready to query.
The data is public. Making it usable takes months.
Federal road-safety data is split across 50 years of changing formats, codes and files, and recalls are published separately. Making it usable takes months of engineering before your team can ask its first question.
We've done that work once. You start at the query.
Four record types, linked and ready.
Crashes, vehicles and people join on shared keys. Recall records ship in the same dataset.
Crash records, 1975 to 2024, on one normalized schema.
Vehicle records, each tied to its crash.
Person records, each tied to its vehicle and crash.
Recall records across 30,438 campaigns.
Three industries, one dataset, different questions.
Automotive, insurance and public-sector teams work from the same linked records. What changes is the question.
Automotive and mobility
Compare vehicles on real-world fatal-crash outcomes and study the scenarios driver-assist systems have to handle.
- 01Vehicle safety benchmarking
- 02Recall intelligence
- 03ADAS scenario research
Insurance and fleet
Add 50 years of road-safety context to research, fleet programs and the risk models you already run.
- 01Geographic safety intelligence
- 02Fleet safety programs
- 03Enriching risk models
Transportation and government
Find where fatal crashes concentrate, decide where to spend, and check what changed after a policy took effect.
- 01Crash hotspot screening
- 02Prioritizing infrastructure spend
- 03Pedestrian and cyclist safety
- 04Evaluating policies
Questions you can answer on day one.
If your team is asking something like this, the data is already shaped for it.
Which vehicle model years show the most rollover fatalities?
Where do pedestrian deaths cluster at night?
How many recall campaigns cover a given component, by manufacturer and model year?
How do fatal crashes involving large trucks vary by time of day and road type?
Which counties have the longest gap between crash time and EMS arrival?
How does alcohol involvement in fatal crashes shift by hour, weekday and state?
Which states see the highest share of fatal crashes in rain or snow?
How has the number of cyclist deaths changed by state over the last decade?
Why not build it yourself?
You can. Here is what that takes, next to what you get on day one.
Official sources, every step documented.
Built only from NHTSA data. Nothing is estimated or blended in from other sources.
FARS
Fatality Analysis Reporting System. The source for every crash, vehicle and person record, 1975 to 2024.
Recall flat files
The source for all 328,859 recall records across 30,438 campaigns.
Every transformation documented
You can trace any field back to the source file it came from.
Original source codes kept
Normalized values sit next to the codes NHTSA published, so nothing is lost.
Totals checked against NHTSA
Counts match NHTSA's official numbers for 2010–2024.
NHTSA and Google do not endorse this product.
How access works.
One linked, verified dataset. Access by arrangement only.
- 01
Request access
Tell us who you are and what you want to answer.
- 02
We arrange access
We confirm terms and set up access for your team.
- 03
Start querying
Use SQL, your notebooks and your BI tools against the linked dataset.
- Delivery
- Linked, query-ready dataset
- Access
- By arrangement only
- Release
- One-time verified release
- Refresh
- No automatic refresh
It's the data, linked and verified. If you need analytics or applications on top, we build those separately.
The limits, stated plainly.
What the dataset covers, and what it doesn't.
Does it include non-fatal crashes?
No. The dataset covers fatal crashes only.
Can insurers use it for pricing or underwriting?
We make no pricing or underwriting claims. Insurance use is for enrichment and research only.
Does the data show that a recall caused a crash?
No. Recall and crash records sit in the same dataset so you can study them side by side. Nothing in it establishes that a recall caused a crash.
How current is it?
Crash data runs from 1975 through 2024. This is a one-time verified release with no automatic refresh.
Is there an API, dashboard or SLA?
No. It's delivered as a dataset. There's no API, dashboard, real-time feed or SLA.
Can you build analysis on top of it for us?
Yes. That's a separate engagement, scoped to the questions your team needs answered.