. WHAT TO DO TO GET A RISK ASSESSMENT MODULE THAT WORKS? (Presentation of a disruptive solution)



Insurance companies now find themselves in front of a potential change in business model, and the so-called "pay how you drive" pricing is an idea whose time has come. Indeed, since actuaries analyze their customer profile, projecting accidents on groups of customers, they generate pricing taking into account number of accidents and cost of accidents (risk and loss). Then why not consider a unitary manner to “measure” the behavior of a driver and his/her rick of accident, in real time ? This would adjust the rate, not only by the group to which the driver belongs to (eg youth, seniors, professionals ...) but also by its actual driving behavior NOW (at every second). The challenge is to enable the sorting of policyholders, so as not disadvantaging a careful driver that would be unfortunately projected in a group of risky drivers (because he is young and owns a red car, for example) and do not miss a dangerous driver who would luckily be projected in a safe drivers class (because it was the right age and doesn’t drive much, for example) : Moreover, we think that some companies using "bonus-malus" should make the difference between a dangerous driver who has been lucky (and we all know, it will not last forever) and a careful driver who has had two unlucky accidents last month. It is then important to understand that an accident is defined as a statistically rare event, and we dare say it is a shame to wait for the repeat of dangerous behavior of a bad driver until it leads to accident (knowing that hitherto, he has avoided the accident, thanks to the reflexes of other road users). The risk taken by such a driver is much higher than the insurer thinks. And actuaries compute their pricing taking a wrong risk into account. The overall objective of a better unitary risk assessment is multiple : . gain profitability by keeping good seniors, taking the good youth and deterring dangerous drivers . teach drivers so they collectively reduced the number of accidents and loss (if the risk measurement system allows this explanatory function) . increase customer satisfaction by decreasing injustice For those reasons, and since it is practically feasible, insurers have started to put telematics devices into cars, and they already measure drivers behavior (electronic device may be the smartphone, either telematics boxes, or both together). Ordinary telematics Companies have extensive experience of App (smartphones) and telematics devices deployment. They all started with geolocation devices, which is a very simple application of GPS. Then, they developed software to measure acceleration and braking, and some of them propose estimation of eco-driving (based on estimation of the conservation of the inertia of the vehicle). Eco-driving is important for fleet managers that wish to reduce fuel budget. While these applications may include niceties (gearbox ratio engaged and engine speed, transported mass, etc.), it's still a mathematically simple application, which deals with some thresholds on accelerometer data. Recently, insurers said they are interested in embedded onboard risk estimation (heart of their business), and have turned to their historical partners of telematics to find solutions. These operators believe (or they “want to believe”) that their accelerometer thresholding software can also be used to estimate the risk of driving, and so they have proposed various and varied scores, often based on a measure of the severity of the braking and also on lateral acceleration (speed on curve). NB: these companies have never worked on driver assistance systems for automotive manufacturers or for road safety teams responsible for the development of roads in order to reduce accidents, and however good they are on telematics, they are not experts in complex applied maths solutions for onboard safety measurement. And recently, we could read the first feedback from some Insurance companies (see 7 - REFERENCES), which demonstrate that the measures and estimations made by the historical providers of ordinary telematics devices have absolutely no predictive power regarding the accident ! No actuary in the world, until now, could find any correlation between their accident database and all scores measured in vehicles on the basis described above (severe braking detection, …). We explain in this article why this lack of conclusive results is absolutely normal, and we show that simplistic assumptions made by telematics companies and insurers themselves have no foundation. Finally, we present as an alternative a totally disruptive solution, which has been developed and tested, validated during the last 15 years, and that is now available.

To read the full position paper :