A risk index is the number of driving events in a region, normalized by exposure. Insurers and autonomous-vehicle developers asked for exactly that: a way to differentiate risk per region that updates, rather than resting on static historical factors that do not change over years.
Empiric Earth Risk Index is powered by more than 100,000 cameras active every month. That number is the sensor base. The rest of this piece is about the denominator.
Why it matters
Underwriters price risk on claims history and demographics. Both describe what has already been paid out. Neither describes what a road is doing this month, which is why a region can drift for two renewal cycles before the loss data admits it.
An index built on live observation of driving fills that gap, on one condition: the events it counts are meaningful, and the denominator it uses is honest.
What a risk index is
The events counted are IMU-based hard braking, cornering and acceleration, plus verified collisions that the network annotates and confirms.
The denominators available are miles driven, hours driven, or cameras deployed in a specific region, whether that region is a state, a county, or a city. Which denominator you pick changes the answer. A rate per mile is different from a rate per hour is different from a rate per camera. That difference belongs in the methodology, not in a footnote.
The default view shows the last month, with month-by-month trend history behind it. Risk can be calculated separately for daytime and nighttime driving. A combined figure hides the half of the day where the difference usually lives.
How it works
A rideshare driver is recognizable from a signature of many short local trips inside a defined region. A truck driver covers long runs on interstates. A personal-vehicle driver drives less than either. Those are the observable patterns Empiric Earth reads off the driving itself, rather than from a form somebody filled in.
Below the city-level rate, a heat map plots risk over Uber's H3 hexagons, the standard grid that divides the map into equal hexagonal cells, rather than Google's S2 cells. Heat maps show which zones inside a city carry the most risk.
Where Empiric Earth fits
Autonomous-vehicle developers usually want the raw exposure data rather than the score itself. What matters to them is comparing autonomous safety against human driving, which needs the denominator more than the ranking, and understanding the type of risk and the type of events per mile they should expect before deploying to a new region.
Only we know
Sorting drivers by mobility signature is possible for a company measuring the driving directly. A rideshare driver produces a distinctive pattern of many short local trips inside a defined region, and a truck driver produces a distinctive pattern of long runs on interstates. Neither signature is knowable from an insurance form. Both are visible in the exposure data, which is Empiric Earth's own measurement.
Limits
The current index is a relative ranking of US regions against each other, not an absolute risk figure. A region placed above another is riskier in the ordering. It is not "twice as dangerous", and any sentence that says so has left the evidence behind.
Weather is not yet an input. Time of day is the currently active external factor, and day-and-night can be scored separately.
FAQ
Q: What is a driving risk index? A: A risk index is the number of driving events in a region, normalized by exposure such as miles driven, hours driven, or cameras deployed.
Q: What events does Empiric Earth Risk Index count? A: IMU-based hard braking, cornering and acceleration, plus verified collisions.
Q: How is the score built at different geographic levels? A: State, county or city, with heat maps over H3 hexagons below city level.
Q: What does the risk score not tell me? A: The current score is a relative ranking of US regions, not an absolute risk figure.
Q: Why do AV developers care about the exposure data more than the score? A: Because comparing autonomous safety against human driving needs the denominator, and understanding the events per mile in a region matters before deploying there.