What Is a Driver Risk Score?

Most fleets already know which drivers had a collision last year. A driver risk score tries to answer a more useful question: which drivers are most likely to have one next. It does that by turning everyday driving into a number that can be compared across drivers and tracked over time, long before a crash record exists.

The score itself is simple to read. What sits behind it is not. Two scores can look alike and mean very different things depending on which signals go in, how they are weighted and whether the model accounts for the situation each behavior happened in.

What goes into a driver risk score?

A driver risk score is built from signals collected while the vehicle is being driven. They usually fall into four groups:

• Driver behavior: phone use, glances away from the road, signs of drowsiness, seat belt use.
• Vehicle dynamics: speed, harsh braking, harsh cornering, rapid acceleration and following distance.
• Traffic: the vehicles, pedestrians and cyclists around the vehicle at the time.
• Context: road type, weather, time of day and other conditions that change how risky a given action is.

Telematics-only systems can see the second group but not the first. Seeing where a driver is looking, or whether they are starting to nod off, needs a camera facing the driver, which is the job of a driver monitoring system.

Event counts versus predictive scores

The simplest scores count events: every harsh brake or speeding alert takes points off. That is easy to explain but treats very different situations as the same. A glance at a phone while stopped at a red light does not carry the same risk as the same glance at highway speed while following too closely.

A predictive risk score weighs each signal by how much it actually raises the chance of a collision, and by what else was happening at the time. Two moderate signals that occur together, such as distraction and a short following distance, can add up to more risk than either one alone. The aim is a score that ranks drivers by future collision risk rather than by how many alerts they triggered.

How to read a driver risk score

Scoring systems use different scales and directions, so the first thing to check is whether a higher number means safer or riskier. On the Empiric Earth VERA Score®, a higher score means lower collision risk.

A single number is only a starting point. A useful score also shows which behaviors are pulling it down, so a manager can tell whether a driver's risk comes mostly from inattention, from traffic violations such as speeding, or from aggressive handling. That breakdown is what turns the score into a coaching plan.

Movement matters as much as the level. A driver whose score has improved steadily over several weeks is in a different position from one who sits at the same number but is getting worse.

Does a better score mean fewer collisions?

A score is only worth tracking if it relates to real outcomes. In Empiric Earth data, drivers with a VERA Score® below 20 averaged 7.3 collisions per million miles, drivers around 50 averaged 4.5, and drivers at 90 or above averaged 1.5. A large improvement in a driver's score is associated with a substantial reduction in collision risk.

That relationship is an association, not a guarantee for any single driver. It is strongest when the score is used across a whole fleet and over time.

How fleets use driver risk scores

Fleets use driver risk scores in four main ways:

• Prioritizing coaching. Safety teams have limited time. Ranking drivers by predicted risk, rather than by alert volume, sends that time to the drivers and behaviors with the greatest safety impact.
• Tracking whether coaching worked. A consistent score gives a baseline. If a driver's score does not move after coaching, the approach needs to change.
• Recognizing safe drivers. Scores make good driving visible as well as risky driving, which supports recognition programs.
• Showing risk to insurers. Observed driving behavior gives an insurer more context about a fleet than a loss record alone.

What makes a driver risk score trustworthy?

A score is only as good as the detections that feed it. If the in-cab system misreads behavior or raises false alerts, the score inherits those errors, and drivers stop trusting both. Accuracy of the underlying detections, ideally tested by a third party, is the first thing to check.

The second is explainability. Drivers and managers should be able to see why a score is what it is, in terms of specific behaviors, not a black-box number. The third is consistency: the same driving should produce the same score, so that changes reflect real changes in behavior.

Driver risk scores and privacy

Scoring drivers does not require a manager watching a live feed. On the Empiric Earth Fleet Safety & Operations Platform there is no constant monitoring and no live streaming. Alerts and video workflows are tied to defined safety use cases, and the score is built from those safety-relevant signals.

How Empiric Earth scores driver risk

The VERA Score® (Visually Enhanced Risk Assessment) is the predictive driver risk score in Fleet Intelligence on the Empiric Earth Fleet Safety & Operations Platform. It works in four steps. It measures more than 20 real-time signals across the driver, vehicle, traffic and surrounding context. It scores them as one consistent predictive risk score. It explains the score by organizing contributing behavior into four categories: high risk, inattention, traffic violations and aggressive driving. And it points self-guided or manager-led coaching at the behaviors with the clearest safety impact.

See Driver Risk Scoring for how the VERA Score® fits with Driver Behavior Alerts and Impact-Focused Coaching, or browse related terms in the Empiric Earth glossary.