A Prediction: Data Integration Will Improve Safety

From Tucson to Tucumcari, Tehachapi to Tonopah: As the Little Feat song says, truckers are willing to drive every kind of rig that's ever been made. Over the years, trucking and logistics companies have invested considerable sums in asset management and tracking systems, pushing the envelope with global positioning system (GPS) technology. Recognizing that loads will only go as far as drivers can take them — that accidents hurt people, damage reputations and lead to legal and insurance problems — leading companies are now intent on leveraging information to improve driver safety.

"When you make your company safer, you improve the quality of life," says Reggie Dupré, CEO of Dupré Transport, headquartered in Lafayette, Louisiana. "Rather than reacting to bad things, more of your energy is devoted to things that make your company better." Although the company handles many kinds of loads, it specializes in hazardous material (HazMat) and bulk liquid commodities. Thus, Dupré knows that safety is a big factor in sustaining business growth.

Dupré Transport uses Eaton Vehicle On-board Radar (VORAD), which can warn drivers of potential hazards so they can take immediate evasive action. Using GPS, VORAD also gives the company data about quick stops, sudden acceleration, deceleration and other behavior so that it can work with employees to cut down on dangerously aggressive driving. The company uses Qualcomm fleet management solutions and telematics services that relay information back and forth about the location and security of trucks and trailers. Previously, Dupré Transport didn't store GPS data, but it realized that this resource could help it develop an information-based approach to reducing hazardous driving behavior.

The company's dogged pursuit of safety led to a contract with Circadian Technologies, a research and consulting firm with expertise in extended-hours operations (see Driving Force, below). "We trained our teams, including drivers' families, with our knowledge about fatigue, alertness and how the body functions," recalls Dupré. "But as good as Circadian has been, it's a functional tool." In February 2005, the company began working with FleetRisk Advisors, a risk management and loss-control services provider to determine what it could discover by integrating data from GPS and other the disparate sources. "We collect a lot of data but unless we pull it together, we can't really get close to a predictive understanding of what we still need to work on with our drivers, customers and equipment maintenance," says Dupré.

The first task was to bring the data into a data warehouse: that is, FleetRisk's Transportation Risk Analytics Center (TRAC) running on Oracle servers. FleetRisk provides a universal adaptor that connects all onboard data collection devices with its GPS telematics monitoring system. "[The adaptor] translates all the data into a standard format before it gets imported to the data warehouse," says FleetRisk chairman Sam Wilkes.

With the data integration architecture set, FleetRisk's objective was to develop three risk signatures covering driver fatigue, aggressive driver behavior and theft and fraud. Using aggregation analytics, FleetRisk reported that Dupré Transport's newest drivers were less likely to have a "loss-causing incident" than experienced drivers. "We used this information to reset how frequently we trained drivers and recertified them for HazMat loads," says Dupré.

A risk analytics process offered limited predictive understanding of driving habits and how they relate to increases in the number of incidents. Drivers with hard braking or speeding habits, as recorded via GPS telematics, were "twice as likely to be involved in loss-causing accidents," Wilkes explains.

For more sophisticated predictive modeling and analysis, Wilkes turned to new partner Valen Technologies, which offers Risk Manager, a predictive insurance underwriting software system. While increasingly rules-based, most insurance underwriting functions are still largely dependent on human judgment. Valen instead applies artificial intelligence in the form of computational learning technology to evaluate risk factors and come up with an underwriting risk assessment that strategic business managers use to set economically oriented objectives.

FleetRisk let Valen's algorithms loose on the TRAC data warehouse to discover patterns that showed a driver's likelihood of having an accident. The process used seven years' worth of driver and accident information. FleetRisk used the first six years of data to develop a "blind test" for the predictive software to see if it would offer results close to the actual data for the seventh year. "And in fact, it did," says Dupré. "The group of drivers ranked as the riskiest had more than three times the number of accidents and incidents as the group ranked as least at risk."

Dupré Transport plans to implement a series of predictive models, including some that will cover customer risk and profitability. "If we find that we have accidents in and around a customer's location," says Dupré, "we can see if it's the driving or something that the customer is doing. We can then work with customers to see if we can make things safer."

"I'm as excited about this as anything I've ever been excited about," says Dupré. "When you integrate the data, the leverage you gain is huge."

Integration and Analytics Drive Safety Improvements

At Dupré Transport, GPS telematics data is integrated with other disparate sources to create an information platform for its analytic activities. FleetRisk Advisors' standard interfaces transform the data before it gets into the data warehouse. Once there, FleetRisk's analytics develop aggregated views for insight. More sophisticated predictive understanding comes from Valen's computational learning algorithms; these look for patterns that deepen understanding of "risk signatures" regarding fatigue, driver behavior and theft and fraud.