The US 2025 Traffic Events dataset brings over 104 million telematics-grounded records into a road-segment-level risk scoring layer built for UBI underwriting and commercial fleet analytics. For the first time, your models can price the exact intersection, curve, or on-ramp where risk clusters, not the territory it happens to fall inside.

Most data vendors sell you a national license to millions of records you will never contact. BigGeo works differently.You tell us where you sell, and we cut you exactly that geographic slice of the dataset. A single city.
A cluster of ZIP codes. A metro area. A county. Whatever matches your territory. The result is a lean, CRM-ready file containing only the companies in the markets you actually work, delivered at a fraction of the cost of a full national license.

Usage-based insurance promised precision pricing. What most carriers got instead was telematics data feeding the same coarse territory models that existed before GPS existed. The signal is in the road. The models are in the zip code. That gap is where margin leaks.
Hard braking, sudden acceleration, and aggressive turning events are classified and geolocated, giving your models a behavior signal that accident reports will never provide until it is too late.

Each event carries timestamped coordinates and speed or acceleration magnitude, so you can weight events by severity and build risk scores that reflect how bad a behavior was, not just that it happened.

Events are joinable to the road network at segment level, letting you build heatmaps by intersection, curve, or on-ramp and flag specific corridors for premium adjustment or fleet routing avoidance.

Video availability flags on qualifying events let claims investigators and driver coaching teams surface the footage that matters without manually scrubbing hours of recorded data.

BigGeo AI is live in ChatGPT today and shipping in Claude, giving your analysts direct plain-language access to 104 million traffic events without a GIS specialist or a data pull request. Ask which intersections in a target market generate the most hard-braking events, which corridors should trigger premium surcharges, or which routes a commercial fleet should avoid entirely. The underlying data never leaves the governed compute path, regardless of how the query is structured.
The dataset captures GPS-geolocated events with sufficient coordinate precision to join to individual road segments, including specific intersections and curves. Inside BigGeo's DataLab, your team can join this event layer to your own road network reference data or to other spatial layers without exporting anything. The compute engine reads only the geometry your query touches, so intersection-level queries run at the same speed as corridor-level ones.
Your internal telematics gives you your policyholders. This dataset gives you the road context those events are happening inside. Joining your event stream to a 104-million-record external traffic events layer lets you validate your own risk scores against a broader population signal and identify high-risk segments your own fleet has not driven yet. That is the difference between pricing your book and pricing the road.
The dataset is published as a 2025 current-year vintage. Refresh cadence for this specific dataset should be confirmed at the time of your sample request, and BigGeo surfaces that information at the dataset level inside the Marketplace so your team is never working with assumptions. If freshness is a hard requirement for your UBI pricing cycle, that is exactly the conversation to have when you request a sample.
No geospatial pipeline required. Once the dataset is activated in your DataLab environment, your data science team queries it directly using the tools they already use, or through BigGeo AI in plain language. Variants govern exactly what each role can see and do with the data, so you can give analysts full access without worrying about ungoverned exports. The heavy spatial lifting happens inside the compute layer, not inside your infrastructure.
Request a sample scoped to your target markets or a specific risk corridor you are already monitoring. That gives your analysts real data to run against your existing model inputs before any commitment. Most teams have a working proof of concept within the first week of access. Book a 30-minute call and we will scope the sample to your use case before it lands in your environment.