Predictive Analytics is nothing new, but the application of predictive techniques has been seeping into every facet of business and government operations. The market for predictive analytics technologies tripled, from 11 billion to 35 billion dollars, from 2000 to 2012. It is now estimated that the US alone will need 190,000 more analytics experts and 1.5 million more data-literate managers by 2018. These figures are derived from FICO analytics.
In the past few years, however, even predictive analytics is progressing. It has given way to prescriptive analytics. This change is more than wordplay, introducing a more proactive approach to data analysis. For fleet safety decision-makers, a prescriptive approach introduces a customized process for effectively identifying and remediating hidden high-risk drivers within a specific organizational culture.
Fleet Application of Analytics
Prescriptive analytics in fleet management helps take some of the guesswork out of budgeting for repair costs and allows fleet managers the opportunity to work with drivers towards accident prevention. Managers and drivers can work together to improve safe driving techniques. This has an effect on driver behavior that leads to better driver retention rates.
Preventing accidents is more possible than ever, from a fleet risk management perspective, when a quality prescriptive model is used in conjunction with telematics data and training is provided for at-risk drivers. The data an analyst can collect from prescriptive modeling is called actionable data; the results show patterns that can be addressed immediately with training. Once training is assigned, alerts can be set to show that a driver has completed their training, creating a positive feedback loop between drivers and fleet management.
There is an abundance of predictive models in use across the fleet industry as of late, but whether these models can evolve to a prescriptive level will vary from fleet to fleet. In addition, the specific predictive model used, and its resulting effectiveness, will vary across companies, but most follow a fairly standard set of criteria. The majority of predictive models use MVRs (motor vehicle records), accident history, and traffic violations to create a risk profile for each driver within a fleet. Risk profiles most often assign a risk score to each driver based on their driving history, and most fleets categorize their drivers within multiple risk categories from safest to most at-risk.
How to Apply Prescriptive Analytics to Your Fleet
These prediction models still tend to trend the last three years of records for a fleet’s risk profile, but prescriptive models that track the last five years of available data are being implemented.
The higher the volume of data collected the more accurate the modeling, and newer prescriptive models looking back as far as five years through MVRs and incident reports prove to be most accurate. A five-year model has not been used to assign a risk level that can be used in a punitive manner; the purpose of a deeper, prescriptive model is to gain further ability to assign training to risky drivers that lead to accident prevention.
Effectiveness of a Five-Year Prescriptive Model
A sales client of CEI with 1,489 total vehicles, whose drivers log an average of 1,963 miles per month, used our prescriptive model for one year, and the results show a predicted number of 357 accidents for the year (24% accident rate) and the actual number of accidents for the year was 375 (25.2% accident rate). According to the AAA Foundation for Traffic Safety, the national average for miles driven per month ranges from 888- 1,095 miles. So, for a fleet whose drivers rack up nearly 900 more miles driven per month than the national average’s high end, a difference of 1.2 percentage points from predicted to actual accidents is exceedingly accurate. Also, the number of accidents becomes understandable for fleet managers from a risk management perspective, due to more time spent on the road.
Another client of CEI, with a service truck fleet of 8,855 vehicles, received our estimates 10 months ago that they would likely have 1,926 accidents within the next 12 months. CEI projects, based on the first 10 months of actual accident data, that by the end of the 12 months they will have had 2,155 accidents. That leaves us with a 2.5 percentage point difference from the 21.8% projected by predictive analytics to the 24.3% actual accident rate that we projected. (We are not provided with mileage from this fleet, and they are our first service fleet to test out our predictive analytics model, which explains the incomplete data set.)
The CEI Group’s traditional predictive analytics model has been able to help fleets reduce accidents by up to 35%, and with added knowledge of high-risk drivers from the five-year prescriptive analytics model, we expect that percentage to increase once fully implemented.
Prescriptive analytics success can be measured in two ways. The first success metric is to see how accurate the model was in total accident prediction during the projected time frame. Even better, and this is really the crux of prescriptive analytics, is to look into whether or not any training is given to high-risk drivers, and if, after training, the actual number of accidents is lower than accidents from previous years and the predicted number of accidents.
Many drivers are at first skeptical about being closely monitored with telematics and prescriptive analytics. The opinion begins to shift when a driver is found not at fault in a collision due to irrefutable telematics data, or when a driver is given a training module and can note a difference in their driving behavior.
This is an exciting time in the Fleet Management world, and in business as whole, because of the emergence of new and improved technologies that save time and money in easily identifiable ways. All of these new technologies, including telematics and prescriptive analytics, are data driven, and they are often referred to with the umbrella term “big data.” It is time for all members of the fleet world to embrace big data, telematics, and prescriptive analytics in order to reap the rewards of saving time, money, and the lives of drivers.
By Kevin Reilly, Editorial Communications Manager, The CEI Group, Inc.