Insurance claims are a loss for both, the policyholder as well as the insurance company. For policyholders, filing an insurance claim often means that they had a bad real life experience. For example, a damaged car or a health issue. Moreover, filing claims will jeopardize claim-free years, which lead to higher premiums for the insured. On the other hand, insurance companies incur costs when they have to compensate for these claims. Therefore, it can be argued that it would be beneficial for both parties if insurance claims could be prevented. By combining the power of a predictive risk model with recommendation systems, we can create a tool that generates an outstanding customer experience as well as a positive business impact for the insurer.
'A clever person solves a problem. A wise person avoids it'.
(Albert Einstein, 1878 - 1955)
Insurance companies are in possession of big data sets with information about thousands of policyholders, containing e.g. demographic data, car data and behavioral data from external sources such as navigation systems. These data can even be enriched with external data (e.g. weather data or data about criminality in particular neighborhoods). Data science algorithms can reveal patterns in the data of different policyholders, in combination with their claim history of their car insurance policy. These patterns can be used to create a powerful predictive models that reveal high risk factors and predict the probability that a particular claim occurs for a specific policyholder. To illustrate different risk levels between individuals, I’d like you to introduce our policyholders Thomas and Barbara, and their individual probability of filing a claim on their car insurance policy.
Thomas, a 22 year old construction worker, drives a brand new Audi A3. Data from his navigation system reveals he likes to drive fast, to spend less time commuting to and from his work. These data combined with his claiming history, indicates his speeding behavior can be categorized as high risk behavior. His driving style is not only getting him tickets, it is also an enormous risk factor.
Barbara, on the other hand; a 38-year-old stay-at-home mom, owner of a Renault Espace, only drives relatively short distances. She doesn’t spend a lot of time in traffic, and when she does she has a decent driving style. Based on this data and the fact that she has never filed a claim, it can be assumed that the individual risk of Barbara will be significantly lower than Thomas’.
Now we have the power to identify risk factors on an individual level, we have the knowledge that can be used to prevent future insurance claims from happening. This can be achieved with intelligent recommendation systems. These systems will nudge the behavior of your policyholders to low risk behavior, by recommending small but important changes in your customer’s daily life. To illustrate this, we get back to Thomas and Barbara.
The data coming from Thomas' car indicates that his brakes need maintenance in the near future. However, the data of the insurance company shows that the status of the current brakes in combination with a driving style similar to Thomas' are an high risk factor. Whether, for more decent drivers it has never lead to a problem. To prevent Thomas from getting in a car accident caused by his driving style in combination with the status of his brakes, the insurance company will recommend Thomas to go the garage earlier than the regular periodical maintenance, which was set to be in six months. Besides, the insurance company can provide a substantial discount, to motivate Thomas. This will decrease the chance of Thomas getting in an accident, and is significantly more cost effective than the paying for a car repair.
This could have been prevented with a wise recommendation system
Another example is in case of an upcoming hail storm in the area where Thomas' is currently driving. The insurance company is aware of this fact by tracking external weather data and combining it with geographical information from Thomas' navigation system. Since this increases the probability of car damages, it would be smart if the insurance company informs Thomas in advance. This could simply be done by sending a text message that warns him about the hail storm. The insurance company recommends him to take shelter, to prevent his car from getting dented. This way, Thomas doesn’t get his car damaged, and the insurance company won’t have to pay to repair the damages. You can imagine how happy both parties are with this little piece of advice.
The recommender system is also beneficial in a health insurance context. Unfortunately, Barbara has a high risk at getting diabetes. She is a heavy smoker and has an extensive medical claim history. Lately she is having problems with her kidneys, which is an indicator of future diabetes. Her insurance company invites her over for a free consultation with their company doctor. The doctor recommends making certain small changes in her eating and exercise patterns, to decrease the chance of Barbara getting diabetes. Eventually the insurer profits from this by preventing larger claims for the treatment of diabetes.
Both situations are perfect examples of how tailor-made recommendations can create win-win situations for both the insurer and the insured.
We end up with a complex system with a predictive power that creates a win-win situation for both insurer and insured. Policyholders will experience less bad real life experiences, because of the personalized recommendations. This personalized customer service will make policyholders more loyal, and equally important decrease claims. This in combination with the lower costs for insurance companies, it will finally result in a higher customer lifetime value and a healthier claim/premium ratio for the insurer.
Besides, the healthier claim/premium ratio enables the insurance company to set lower premiums for their policyholders or compensate their policyholders, which give them a competitive advantage to outperform competition.
You might think: what about privacy? And you should, as privacy concerns are getting more relevant in this digital age. Clients should be able to share data and not be afraid of ‘Big Brother’ watching them. They don’t want to pay a higher insurance premium just because they happen to post a lot of pictures of them eating pizza on Instagram.
Key in data ethics is transparency in communication, data sharing and communicating efforts in data security. Voluntary cooperation of the clients should always be the case to secure the trust of policyholders. All in all, is it safe to say there are multiple ways to achieve safe data sharing. We will go into further detail on this in an upcoming article. So be sure to subscribe to our newsletter and follow us on Linkedin.