BY CHRIS VILLELLA
Analytics has spread into every element of business and our daily lives, and the fleet industry now has tools that provide insight into your fleet’s risk exposure like never before.
There are three tiers of analytics: descriptive, predictive, and prescriptive. But, what do these terms mean at their core and to the fleet industry in particular? Descriptive analytics is the most basic, aggregating large sums of big data and breaking it down into digestible chunks of information. Most fleet reports are descriptive in terms of miles driven, accident costs, speeding violations, and gas spend, among other things.
The next level of analytics sophistication is predictive analytics. By using statistics, machine learning, and data mining, predictive analysis determines the probability of something happening in the future.
Prescriptive analytics is the most sophisticated form of analysis. It takes predictive analytics to its natural conclusion: Providing a recommended course of action to reduce the predicted outcome based on expert consultation.
Interestingly, the fleet industry has been using prescriptive analytics for years to complete such tasks as route optimization and maintenance scheduling. Today’s pertinent applications might involve taking a different route in response to an accident or notifying fleet managers when the best date for a tire rotation may be to save fuel. Now, truck fleets can use this technology to enhance their safety programs.
HOW THE PROCESS WORKS
Prescriptive analytics allows fleets to pinpoint those drivers in their operation who are most likely to be involved in an accident and prescribe a personalized regimen of training that will help avoid future accidents. The system works by collecting a broad array of MVR, accident history, and demographic information. Five years’ worth of data ensures the most accurate results.
Dr. Feng Guo, Ph.D. in Statistics and Transportation Engineering from the Virginia Tech Transportation Institute, worked with the CEI analytics team to develop this first-of-its-kind model. After a year’s worth of data, the deeper look achieved through predictive modeling provides insights that are amazingly accurate. Drivers who might have been deemed low-risk using traditional models that rely on three years of limited data are revealed to be high-risk based on five years’ worth of data and assorted other demographic information.
If a driver has not had an accident in the last three years, a traditional assessment will tend to categorize that driver as safe. However, the prescriptive model will show that the driver had a string of rear-end collisions four years ago, for example. History often repeats itself when lessons are not learned, and skills are not taught to combat issues at the root.
CEI looks at the top one percent and five percent of drivers that are predicted to have an accident. Its data shows that drivers in the top one percent and five percent of at-risk drivers made up 4.2 percent of the driver population, but were found to have gotten in 20.1 percent of the total accidents. The CEI data analytics department looked at more than 85,000 drivers for these statistics.
15 percent of the 85,000 drivers have more than one accident on their records, and this includes almost 90 percent of drivers in the top one percent of at-risk drivers and 80 percent of the top five percent risk group.
The correlations are clear, drivers that are predicted to have an incident tend to find themselves involved in an event within the allotted timeframe. These accidents are occurring regardless of where the traditional model placed them.
PREDICTIONS TO RESULTS
Depending on the client’s driver policy, a range of remediation activities is possible, including training aimed at specific crash types within the fleet, managerial intervention, behind-the-wheel training, and more.
However, companies can find the right training by looking for trends in the fleet’s accident data as a whole, and prescribe proactive online training to high-risk drivers that will combat those pain points. Behind-the-wheel training is also an effective means of retraining drivers with troubled driving records. Of course, it costs more and takes more time away from your drivers. Still, behind-the-wheel training can be effective on a more individualized basis.
Another remediation alternative that has been deemed effective is managerial intervention that heightens the urgency for behavioral change or situational training. Creating a dialog with a driver can also illuminate the root causes of their driving difficulties and make choosing the best course of training more effective.
The first step to measuring the success of prescriptive analytics is to see how well the predictions match with the actual number of accidents that occur over the course of the year. The next step is to analyze the effects the assigned training had on the number of accidents vs the predicted number.
Measuring the success of a prescriptive analytics program depends on how focused fleet managers are on prescribing the recommended steps for remediation. The predictions and recommended actions can only help drive down accident rates if the remediation for suggested drivers is followed through by everyone in fleet operations.
Insurance companies are building their own prescriptive analytics models after seeing the precision and success that fleet operators are achieving, so change is imminent across all facets of the automotive industry. Fleet drivers hold themselves to a higher standard on the road, and prescriptive analytics provides fleet operators with tools to continue that tradition of excellence.
ABOUT THE AUTHOR:
Chris Villella is the senior director of account management for The CEI Group. CEI offers detail-driven repair management and driver behavior correction. CEI is the only fleet accident management company to be awarded Gold Status by the Inter-Industry Conference on Auto Collision Repair (I-CAR) for excellence in collision appraisal. Find out more about CEI products and services, visit www.ceinetwork.com.