New report: 10% of drivers account for ~50% of crashes

For years, fleet safety has been built around the individual event. A speeding incident gets logged. A harsh brake gets flagged. A phone glance gets reviewed. That structure has worked, but it leaves safety leaders asking the same question: which patterns of behavior actually matter, and how early can they be caught?

New research from Samsara, based on its proprietary Risk Model (patent pending), answers that question. The 2026 Samsara Compounding Risk Report analyzes aggregated telematics data and finds that crash risk doesn't live in any single behavior. It lives in how behaviors combine, and it's often visible weeks before a crash happens. Key findings:

• The riskiest 10% of drivers account for nearly half (47%) of all crashes.

• The riskiest 30% account for more than three-fourths (76%) of all crashes.

• Drivers with combined mobile use, distraction, and harsh braking are 5.4x more likely than average to fall into the highest risk tier.

• When comparing a driver who is subsequently involved in a crash with one who is not, the model prioritizes the crash-involved driver three out of four times.

Let’s dive in.

The problem with looking at one behavior at a time

Single-behavior data is real, but it's diffuse. Drivers who brake hard, use their phones, or speed are all somewhat more likely to be in a crash, individually carrying a 1.2x to 1.8x increase in crash likelihood. Spread across a fleet of hundreds of drivers, that's not a prioritization strategy. It's most of the fleet, most of the time.

Fleet managers already know this problem. Every alert demands a decision: is this driver actually a risk, or is this noise? Single-behavior monitoring forces managers to triage events one at a time, with no way to tell which alerts are worth acting on.

When risky behaviors combine, risk multiplies

Samsara's proprietary Risk Model grouped drivers into profiles based on combinations of driving behaviors and contextual factors, then measured how likely each group was to fall into the highest risk tier. The results show that combined behaviors don't just add risk together. They multiply it.

A driver who uses their phone while driving is 2.7 times more likely than average to fall into the highest risk tier. Add harsh braking, and that number climbs to 4.5x. Add distraction or night driving on top of that, and it reaches 5.4x.

That's the core finding: risk concentrates in the combination, not the component. A pattern of mobile use, distraction, and harsh braking together is a far stronger signal than any of those behaviors seen alone.

A small share of drivers accounts for most of the risk

Crashes aren't distributed evenly across a fleet. They concentrate.

• The top 10% of risk-ranked drivers account for nearly half (47%) of all crashes.

• The top 30% account for more than three-fourths (76%) of all crashes.

That concentration is what makes coaching prioritization the highest-leverage move a fleet safety program can make. By concentrating the most intensive coaching on just 10% of drivers, fleet managers can reach the group associated with nearly half of all crashes. Combine that with a strong safety culture and self-coaching to reinforce safer decisions across the broader workforce, and managers can focus their limited time where it has the greatest impact, without losing sight of everyone else.

The patterns show up weeks in advance

The most useful part of this finding isn't just that risk compounds. It's that it's visible early.

When comparing a driver who is subsequently involved in a crash with one who is not, Samsara’s Risk Model prioritizes the crash-involved driver three out of four times. This holds across both the next-day and seven-day evaluation windows.

For fleet managers, this changes the nature of the safety conversation. The question shifts from "what just happened?" to "which drivers should we focus on this week?" — and the model provides a consistent, reliable answer to that question before anything has gone wrong.

Turn these insights into action with Coaching Priority

Underpinned by the Risk Model, Coaching Priority is Samsara's proprietary framework for surfacing and prioritizing coaching opportunities across your fleet — the only solution of its kind available today. 

The AI-powered feature evaluates ~50 factors, including driving history, weather, and driving behavior patterns, to give managers a single view that highlights where coaching will have the greatest impact, while drivers who are already building strong habits receive automated self-coaching to keep them on track.

“Coaching Priority shows us exactly where to focus—down to the specific locations, behaviors, and drivers driving risk,” said Tom Karnowski, Vice President, Environmental, Health, and Safety at USIC. “That clarity empowers us to take targeted action and elevate safety across our entire fleet.”

Get the full report

Download the full 2026 Samsara Compounding Risk Report to see the complete findings, example driver profile breakdowns, and a full methodology built on causal and statistical analysis.