Skip navigation

Using driver data to predict turnover

if driver data can be plugged into a software-driven model to rank those most at risk of having an accident, the same can be done to rank those with the highest risk of leaving their jobs

The way Dean Croke sees it, if driver data can be plugged into a software-driven model to rank those most at risk of having an accident, the same can be done to rank those with the highest risk of leaving their jobs. That would help to lower the trucking industry’s driver turnover rate, which topped 110% in 2006.

Croke, chief product architect at FleetRisk Advisors, plans to use two-years worth of historical electronic data extracted from multiple sources within a trucking fleet—over 300 data points including everything from the marital status, number of cash advances per month, even tenure at previous jobs. Pattern recognition software will be used to scan the data to identify the precise combination of data elements that contribute to optimal driver retention.

“We want to figure out what characteristics keep a driver at a trucking company, along with the ones that make him leave,” Croke told FleetOwner. “One reason you make the data cache so large is that even obscure data points—such as home residence zip code or length of haul—creates links in a bigger picture as to why a driver leaves or stays.”

Some of the information FleetRisk is looking to use in its model includes logistics data (warehousing, dispatch, dispatcher, freight type, detention), financial data (load revenue, operating margin, revenue per mile), performance-based data (accidents, incidents, service failures), driver data (pay rate, mileage, tenure, experience), and external factors such as demographics, traffic, and weather.

Much of this data would come from dispatch, human resources (payroll, recruiting), insurance & claims, telematics, warehousing (if applicable) and fleet maintenance. The goal is to develop a “target model” of drivers that are safe, productive and remain with the company for over two years.

Croke said the toughest part of the process is building the model itself, which requires four to six months of work. But that once its in place, it would be relatively easy to plug ongoing data into the software to track driver turnover risk. Driver retention levels, training and hiring costs will be monitored as the program is implemented. Croke envisions the tool will be used to increase both the average driver’s tenure at the company and reduce the carrier’s accident rate through the hiring of safer drivers, he said.

“A truck driver’s job is a hard one, so what we hope to do is create a method that ‘scores’ all of a fleet’s drivers against the traits of the very best in the company—the ‘super drivers’—so the fleet can more readily address problems,” he said.

Croke, a driver himself during part of his 20-year industry career, noted using data models like this to manage turnover are already in use. He pointed to one truckload carrier uses a similar modeling system, assigning point values to negative routines such as low-pay deadheading runs or delivering loads in the Northeast. When a driver reaches 15 points, the fleet can stage a turnover “intervention” of sorts so the accrual of negative experiences doesn’t lead to a driver quitting his or her job.

“This is a very proactive approach, but it uses data points the company thinks will make a driver leave—not necessarily what the drivers themselves think makes them unhappy or not,” he said. “A model based on the driver themselves empowers companies to reduce retention and recruitment costs by keeping current drivers happy, safe, productive and employed, rather than spending millions of dollars on trying to find new ones.”

To comment on this article, write to Sean Kilcarr at [email protected]

TAGS: News
Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.