State of the Industry
Green Guide to Risk Management
Eric Shishko, Senior Vice President, Insurance
The next generation of driver risk management and measurement – overcoming the limitations of telematics and exception-based video
The current state of the art for driver risk management solutions consists of solutions that (1) enable and empower behaviour change in drivers and (2) provide other stakeholders the information they require to predict the inherent risk category that driver occupies. Until recently these two requirements have not peacefully coexisted. That is, approaches to these problems that exhibit effectiveness in changing behaviour (i.e. through limited observation) have not provided the analytical foundation to predict the inherent risk of individual drivers. Alternatively, analytical tools that identify historic correlations between past behaviours as indicators of future risk have been of limited value in changing behaviour.
Let’s first look at changing the behaviour of drivers, or simply put, getting drivers to drive more safely. Why? When drivers drive more safely they crash less often and less severely, saving money, time and most importantly lives.
The approach taken by a few of the early technologies was to use short video clips of risky driving behaviour. These clips—typically 10 to 20 seconds of both the driver and the area directly in front of the vehicle—are captured when a significant driving event triggers a set threshold level, causing activation of the video. The images captured often include a brief period before the trigger and then a similarly brief period after the trigger. These images of the driver, captured at the moment of high g-forces on the vehicle, are then typically sent to a location where a human reviews the image and rates the driver’s behaviour. The human also makes subjective judgments as to the behavioural root causes of the event.
The videos are then made available to the driver's supervisor (or parent in the case of the teen) who can use them to encourage the driver to behave differently. As a behaviour modification tool, video effectively changes behaviour in the short term. That is, the behaviour change continues as long as the driver is constantly reminded that there will be a consequence if they trigger the video. This concept is supported by the Hawthorne Effect and as long as the camera is present and the driver is reminded of the threat of capture, they will generally adapt to the programme and try to trigger the camera less often.
Unfortunately, the behaviour change is tied to a negative consequence and the observations comprise only 10 to 30 seconds of a driver's week. This is the primary limitation of the video approach. While a professional driver may be on the road for 30 or 40 hours a week, the information collected on this total amount of driving is limited to the 10 to 30 seconds, during which the vehicle was subjected to some significant g-force, perhaps hitting the brakes or a pothole or taking a turn too quickly. This means that for 99.99+% of the time a driver is on the road, there is no information about the way they are driving.
Economics are the reason for the limited video footage. Recording, transmitting, and reviewing video is very expensive. The amount of data collected (seconds of video) is generally limited to that which is just enough to support the Hawthorne effect and just enough to show very severe risks. The corporate fleet owner or parent of a teen has similar cost factors. In the case of a corporate fleet, the driver must be pulled off revenue producing activities to review the clips, and a person must be assigned to manage these interactions, all of which create a significant overhead.
Other driver risk management approaches (such as early efforts to implement Pay as You Drive (PAYD) programs) have centered on the distance the vehicle is driven or the conditions under which the vehicle is driven. These approaches capture what is often called the “exposure to risk.” Simply put, the more miles you drive (and the more risky condition i.e. snow) the more you are at risk. Few would disagree with this as a fundamental correlation. These measurements may soon guide insurance rates but there are inherent limitations. Exposure many not be something that a driver can adjust. If you have to drive 40 miles to work you have to drive 40 miles to work - what the driver can control is the WAY they drive those 40 miles. Their behaviour is within their control, and therefore can be positively impacted.
The good news is that there have been recent advances in the ability to use analytics to automatically capture all the risky driving behaviors of a driver for the entire time they are on the road. This approach also provides immediate feedback to the driver for each risky manoeuvre, empowering them to understand and change their risky behavior in real-time. By capturing the complete driving picture, not just 20 seconds per week of extreme movement, a total risk picture can emerge. The foundation for this advanced analytics approach is based on extensive research into the causes of crashes and is tied to several major factors including the type, frequency and the severity of risky behavior.
The value of these next generation driver risk management technologies is two-fold. First is the ability to inform and change driver behaviour, and second is the ability to accurately measure the inherent risk of driver based on the patterns of their driving and the nature of their risk.
Corporate fleets and their insurers are increasingly faced with a very competitive, price-driven insurance market. Reducing the risk of crashes is a positive in the short term as accident rates, costs, and insurance loss ratios will decline, and profitability for everyone will improve. However in today’s price sensitive markets, short term economic gains will soon diminish as prices adjust around the new business models of companies that have the ability to accurately predict risk. In effect, the risk predictions will become a part of the pricing models.
Therefore, insurers can no longer afford to build a cushion into pricing to offset less effective risk assessment capabilities or operational efficiencies (CapGemini 2006). As pricing becomes more central to insurance purchases, insurers will have to improve both price matching and risk segmentation. In order to achieve this more granular pricing, insurers will require more accurate risk predictors.
Creating a risk profile for a particular driver involves collecting information that is correlated with and causative to crashes. This will include many factors, and with the new technologies now widely available, driver behavior will increasingly be included in this mix.
The next generation of driver risk management gives fleets and insurance companies the ability to change driver behavior AND accurately measure driver risk based on driving patterns and other external factors. Measuring driver behavior behind the wheel, and the risk contained within the total driving picture are two critical elements now within our realm of understanding, and both will continue to increase in importance as these new technologies evolve.
