Roadways are getting more dangerous, leaving safety directors with a difficult mission: Sift through terabytes of data to identify gaps in fleet safety. As data analytics tools improve, safety leaders have to learn to use those tools to ride the tsunami of data, keep their heads above water, and ensure the safety of their drivers.
“Even if you have figured out what data is there and what you want to do with it, it’s never in the same place. It’s in systems all over,” said Brian Filip, chief technology officer and head of product at safety software provider Idelic during a Truckload Carriers Association webinar on promoting driver safety with machine learning.
“You’re probably running a separate system for ELDs, for telematics, for cameras,” Filip said, “and getting them to talk to one another doesn’t usually happen, so you spend an inordinate amount of time going through all these different places.”
Machine learning software, such as those provided by Idelic and others, can collate that data and present them to management as actionable insights.
See also: Transforming data into intelligence can give fleets the edge
Managers can use the intelligence to identify high-risk drivers, allowing targeted coaching to more efficiently use company resources and time, said TForce Freight’s director of health and safety, Scott Reagan.
“We’re able to identify those drivers … helping everyone—the operation, the local leadership, the driver,” Reagan said, adding, “We need to get ahead of it instead of talking after the crash.”
Driver scorecards are common for predictive analytics safety software. According to Reagan, the scorecards, if not used punitively, also can identify good driving, fostering a spirit of healthy competition.
TForce drivers “all of a sudden wanted that information,” Reagan said. “It was a gamification. … We saw that risk behaviors start to diminish.”
Michael Lasko, a former driver who is now director of EHS and quality for Boyle Transportation, said adopting new technology can aid retention efforts if drivers are included during every step. Technology considered for use by Boyle is beta-tested by select drivers.
“In my driving experience, one of the most frustrating things about being a truck driver: You don’t hear ‘thank you’ and ‘good job’ anywhere near as much as you should,” Lasko said.
Thanks to driver scorecards, exceptional truckers can be recognized, and their efforts can be commended to the entire fleet, Lasko said.
Boyle creates a professional development plan at 30, 60, and 90 days. The carrier’s department heads participate in meetings with drivers to evaluate their development and, according to Lasko, this has made all the difference.
“It creates this environment where they know they’re part of the team,” he said. “Are they getting supported? Do they feel overwhelmed? And we can work to address that with the driver, and it’s made a big difference.”
See also: Fleets moving from preventive to predictive maintenance save more
AI and machine learning: What’s the difference?
As the market progresses, so will the use of machine learning to analyze safety trends. Machine learning and AI are often used interchangeably, although according to Filip, “they’re absolutely not the same thing.”
AI is a broad term to describe making a machine do something that a human would otherwise do, whether that’s a factory watching for defective M&Ms or Siri recommending restaurants., Filip said.
Machine learning, he said, is allowing a machine to learn and solve problems on its own that a human otherwise could not. For trucking safety, event prediction machine learning is made to analyze trends, such as driving patterns, and highlight ones that it deems high risk so safety management can coach the drivers who need it.
Predictive analytics are one tool in the box
Filip did caution that machine learning technology should not be the last word and that fleet’s data insights must still be incorporated into a robust safety operation.
“Machine learning is not magic,” Filip said. “It is not going to be right every time, and just because it is wrong in some instances doesn’t mean it’s not working right.”
Filip likens predictive machine learning to a weather forecast: The forecast is not always correct, but it is a highly educated guess, and the accuracy of forecasts is far better than 50 years ago.
And the benefits of predictive analytics may not be immediately apparent, according to Filip.
“You can’t draw conclusions on what happens in an hour, or a day, or even a week,” Filip said, because crashes are statistically rare, and carriers can’t look at that granular of a level of data. “You have to measure these things over the course of months, quarters, and even years to really see whether or not you’re having a statistically significant impact on outcomes.”
TForce’s Reagan said, despite technological advancements, “The safest piece of technology we have on our vehicle today is an awake, aware driver.”
