Co-authored with David R. Hardoon and cross-posted on www.davidroihardoon.com
Disclaimer: The views represented in this article are the personal views of the authors
The insurance industry dates back to the 17th century BC when the Babylonians developed a system practiced by early Mediterranean sailing merchants. If a merchant received a loan to fund their shipment, they would pay the lender an additional sum in exchange for the lender’s guarantee to cancel the loan should the shipment be stolen, or lost at sea. Despite having come a long way over the last three millennia, and the more recent impacts to the industry from emergence of AI, the underlying fundamentals of insurance have not changed much since the Babylonians’ times.
Today, the field of insurance is typically divided into the two main areas of Life and General Insurance. General Insurance encompasses areas such as Motor, Travel, Fire, Home, Critical Illness, Liability, Workers Compensation, and Health Insurance. In the U.S., General Insurance is also known as Property and Casualty Insurance whereas in Europe it is simply referred to as Non-Life Insurance.
One of the main drivers of change is the availability of new types of data for the calculation of risk. In motor policy risk underwriting, insurance companies have historically considered attributes such as age, driving experience, driving record of the applicant, characteristics of the motor vehicle, current and proposed mileage, storage of car (indoor / outdoor), anti-theft, anti-lock braking systems, among other factors for the estimation of risk. However, all of these attributes save for the first three, are more likely to help estimate the likelihood of the vehicle being involved in an accident rather than the estimation of the likelihood of the driver causing the accident; as these are attributes that do not represent how the driver performs on the road.
The proliferation of internet-of-things (IoT) based telematics devices, in the form of on-board diagnostic (OBD) devices or ML-driven mobile applications for telematics, have made it possible to capture the intricate details of actual driving behaviour (such as acceleration, deceleration, braking, bearing, speed, hard brakes, hard turns) on a real-time basis. These facilitate a more accurate estimation of a person’s risk while driving. Telematics, and the AI analysing this data, has ushered in the possibility of ultra-personalised driver risk scoring, which aims to present a more accurate view of a person. As a result, an 18 or an 80 year old driver may drive similarly and thus have comparable risks.
The collection of real-time behaviour data poses a number of considerations with regard to privacy and the insurance business model as a whole. In traditional motor underwriting, risk and premiums calculations are usually done on an annual basis. Given that the commonly used data attributes do not change frequently, an annual calculation is reasonable in capturing the fluctuations and changes to risk. However, if we are now able to collate both behavioural data as well as real-time driving data, risk can equally be calculated on a near-real-time basis. This allows a more precise measure as risk is increased or decreased.
What would this mean to the business model? Should risk be calculated in such a manner, and should the pricing model be significantly more dynamic in nature? Our view is that it would depend on the context under consideration. For example, in the relatively smooth and organised traffic of North American, European, or major Asian cities such as Singapore, Seoul, or Tokyo, telematics scoring could be a viable business model for insurers to underwrite risk, and could be an attractive proposition for customers to get good-driver discounts. However, telematics scoring that captures second by second driving behavior may not be very effective in gridlocked traffic in South Asian cities like Mumbai, Delhi, Bangalore, and Dhaka. Nevertheless, telematics could still be used in the emerging economies when scoring drivers during weekends and longer trips on the expressways.
In addition to the behavioural factors discussed, many telematics scoring providers are also proposing a variety of new factors for scoring, such as locational risk (types of roads–interstate, city, residential, school and hospital zones), mobile, sms, and application usage while driving (distraction indicators). Some telematics providers are also introducing a social component by gamification of the driving process, issuing best driver badges among friends, and accumulating good driver scores which can be exchanged for fuel, food, drinks, movie tickets and other perks.
The incorporation of telematics and AI in insurance is as inevitable as the adoption of electricity, automobiles, and steam engines in the last couple centuries. Scoring people based on how they drive, and not based on secondary rating factors such as age, gender, or the zip code, could become the norm over the next decade. Nonetheless, it is important that we proactively and continuously assess where the underlying business models need to evolve to accommodate, and address the consequences, arising from these new technological and data possibilities.