In every insurance contract negotiation, there are two main concerns for the involved parties. The insurance provider worries about making an accurate risk assessment, while the customer needs to justify to himself that he’s paying a fair price for the acquired service.
Failing traditional assessment models
For decades now, the insurance companies used to gather and analyse static data, prior to offering a contract. In the health insurance sector, information like age, gender, good or bad personal habits and health records have been the key deciding factors for the pricing. On another front, car insurance providers would base their contract terms on age, gender, years of driving experience, age, brand, model and engine of the vehicle, type and frequency of use, as well as whether it would be driven in a city or a countryside terrain. Nevertheless, no matter how logical the above approaches are, they have been failing to depict reality in a similar fashion.
First of all, both methods rely on single touch point risk assessment. Pricing and premium decisions are based on whatever information is available prior to signing the contract. The only reason why this information will be updated, is when there is a claim from the customer. But how risk-free or fair is this? For example, why should a 40-year-old with no health record, who follows an obese lifestyle and a junk-food-based nutrition, be considered less likely to suffer a heart attack than a 55-year-old who exercises regularly and has a balanced diet? Or why should a 50-year-old lady in the countryside, who drives moderately and maintains properly her 20-year-old hatchback, should pay higher insurance fees, than a 30-year-old man who drives at top speeds in rural areas and neglects proper service of his 5-year-old convertible?
And at the end of the day, could there be a more realistic risk evaluation model, that would suit better the needs of both the insurance providers and their customers?
IoT-enabled Personalized Insurance Products
Fortunately, recent technological advancements have provided the tools for the development of more personalized insurance products. Based on data collected by IoT sensors and analysed by artificial intelligence (AI) algorithms in real-time, insurance companies can keep promptly updated profile of their customers and their assets (i.e. vehicles, homes etc.) and generate holistic reports with undisputable information. Provided that such provision is made upon the contract agreement, when it’s time for insurance fee discounts or premium surcharges, both parties will know where these decisions stem from.
Applicability and benefits of Personalized Health Insurance models
Nowadays, the insurance companies are in position to collect much more information around a person’s health. IoT smart sensors and wearables enable the transmission of real-time data around physical activities, body metrics (i.e. blood pressure, breathing etc), sleep, nutrition and quality of life, so that they can be analysed and correlated by machine learning (ML) algorithms. This helps the insurance agents to generate personalized risk predictions and more realistic health profiles. More so, a customer could now pay a higher fee for repeatedly neglecting their health or be handed a discount for running a healthier lifestyle, in the context of an existing contract.
Applicability and benefits of Personalized Vehicle Insurance models
Similar is the emerging approach of Usage-Based Insurance (UBI) in the vehicle insurance sector, where the most popular models are:
• Pay-As-You-Drive (PAYD), which uses more traditional-type information, as measurement from the odometer and
route identification by the GPS signals.
• Pay-How-You-Drive (PHYD), which uses more behavioural information, such as the frequency and intensity of braking
and accelerating the car.
The above measurements can be continuously monitored by IoT sensors and transmitted wirelessly in real-time via an HMCU (Hybrid Modular Communication Unit) which is installed in the vehicle. This raw data is fed to cloud-based AI/ML algorithms and can be correlated with contextual information (such as weather conditions and traffic incidents) to produce highly personalized, accurate driver profiles. Though the use of APIs (Application Programming Interfaces), this information can be further forwarded to the insurance company’s ERP system, to enable data-driven decision making.
Concerns around Personalized Insurance Models
As in every technological breakthrough, there are a few concerns raised over the appropriate and risk-free use of the above methods. Probably, the most common ones evolve around:
• Usage rights: The collection of real-time health and behavioural data is falling under regulations such as the General Data Protection Regulation (GDPR). Therefore, the insurance agents need to convince their customers to give consent for the collection, storage and use of such information, probably by offering discounted fees from standard ones, for improved health habits and driving performance.
• Fraud detection: Not only do the agents need to prove to their customers that they are being graded based on faultless, objective data, but also, the customers must retain a high level of credibility, too. For example, it may be tempting to change the driver’s identity information in the event of an accident or give a health wearable to another person with a more active lifestyle. That’s where technology – and more precisely artificial intelligence (AI) and machine learning (ML) – can give a helping hand. Through deep analysis, they can reveal abnormal behavioural patterns and alarm the insurance companies for possible deceptive activities.
Present and future of the Personalized Insurance Models
From the above consideration, one can conclude that personalized health insurance products and user baser insurance models for vehicles can provide benefits to both the insurance providers and their customers. On-going research and development on IoT connectivity, smart devices and AI/ML processing models, will guarantee the extended applicability of such methods.
One of the initiatives aligned with this goal, is our H2020 INFINITECH project, which develops novel technologies and innovation use cases of Big Data, IoT and AI solutions in the finance and insurance sector. UBI related data assets and datasets have been developed by INFINITECH partners ATOS Spain, CTAG, Innovation Sprint, and RRD, and can be found in the INFINITECH marketplace and YouTube channel, including:
• A Processed Synthetic Real World Data (RWD) dataset for tristate modelling: This is a model learning dataset, created out of the “Raw Synthetic RWD” raw dataset, which aims at a predicted outcome for the scope of health self-assessment.
• The SUMO Vigo vehicles sample : This is a dataset from a large amount of vehicle data in the Vigo area in Spain, as generated by the SUMO urban mobility simulator. It describes trip data for several vehicles in the area, with each trip composed by several time instants where the vehicle reports its position (coordinates + road ID), speed, fuel consumption, noise level, etc.
• A Personalized insurance product based on IoT connected vehicles: The presentation of this solutions gives a comprehensive idea around the system architecture and the benefits of deploying Usage Based Insurance models in new vehicle insurance products.
To learn more about INFINITECH’s solutions:
• Visit the INFINITECH project web site: https://www.infinitech-h2020.eu/
• Subscribe to the INFINITECH YouTube Channel: https://www.youtube.com/channel/UClVeOyQyljdCpL51GSPa7Zg
• Subscribe to the INFINITECH Newsletter: https://www.infinitech-h2020.eu/contact-us
• Register to INFINITECH Marketplace: https://marketplace.infinitech-h2020.eu/