Fleets Say Data From ‘Predictive Modeling’ Assists Them in Exposing Safety Problems

SAN DIEGO — By pulling together broad swaths of diverse data through a predictive modeling tool, carriers can uncover safety risks that probably would have gone unnoticed otherwise, fleet safety officials said.

A spike in hard-braking incidents, for example, could point to a problem outside of work that’s weighing heavily on a driver and therefore changing that driver’s behavior behind the wheel.

That kind of insight can give fleets an edge in safety, according to representatives from four carriers using the predictive analytics service provided by FleetRisk Advisors, a business unit of Qualcomm Enterprise Services.

“There’s no field in the database for a sick child at home, or going through a divorce, or just filing for bankruptcy, but the driver’s behavior changes due to those stressors, and it’s reflected in the data,” said Chris Orban, senior manager of software application engineering at FleetRisk, which Qualcomm acquired in early 2011.

During a panel discussion June 5 at Qualcomm’s Vision 2012 Management Conference here, Thom Pronk, corporate vice president of recruiting, training and safety at C.R. England, described how the predictive model can help carriers proactively address safety risks.

As an example, Pronk described a situation where the predictive model identified a driver for risky behavior. Having been alerted to this behavior change, a company safety manager talked to the driver and learned that the driver’s son had been in an accident.

With that knowledge, the company was able to take that driver off the road, get him home and reassure him that his job would remain secure, Pronk said.

Apart from likely avoiding a “pretty significant safety risk,” the company was able to help “in a much bigger way” by getting that driver home so he could take care of a challenging personal situation, he said.

When the predictive model indicates a higher risk level, it drives conversations between managers and drivers to resolve the issue, Orban said.

Something like an unforeseen delay also can change a driver’s behavior and introduce greater safety risk, said Al LaCombe, director of safety for Dupré Logistics.

A driver who falls behind in his daily schedule, often due to circumstances outside his control, will tend to rush to get back to the terminal on time, which could lead to hard-brake events and higher speeds, LaCombe said.

By taking actions to address those types of risks, Dupré Logistics has reduced its severe crashes by more than 30% since it began using FleetRisk’s analytics about seven years ago, LaCombe said.

But in the end, the biggest benefit is in keeping the company’s drivers safe and happy.

“The most important asset we have at our company is our employees,” LaCombe said.

FleetRisk’s predictive modeling is also designed to serve as a tool for driver recruitment and retention.

John Walton, director of safety and compliance at Averitt Express Inc., said his company coaches its fleet managers to initiate positive conversations with drivers about risk data by discussing their concerns and challenges.

“The drivers at the end of the day are just excited that you talked to them, that you listened to them, that you cared,” he said.

Doug Schrier, director of continuous improvement at Covenant Transport, said his company has already seen a difference in driver retention since rolling out the predictive model a month and a half ago.

His company’s fleet managers started having conversations with drivers about their concerns, and “I can’t tell you how many of those drivers were on the verge of quitting and are still with us today,” he said.

Norman Ellis, Qualcomm Enterprise Services’ vice president of sales, services and marketing, told Transport Topics that the strength of predictive modeling comes from the variety of information that it takes in.

That data includes operational information, such as the types of loads drivers are delivering and the lanes they’re running.

But the model also uses personal information about a driver, which can include credit history and information from a job application, such as marital status and how frequently a driver has changed jobs in the past, he said.

“At first it’s hard to comprehend how those things can really be that constructive to an analytic model, but they are,” Ellis said.