Identifying trends from millions of data points, commonly referred to as big data, is opening up big opportunities for Aegon, and has been identified as a key area in which the company can support its strategy.
In Aegon Hungary, a higher level of automation has been made possible as the result of a new algorithm. It works by analyzing the detailed history of client behavior logged by the company's database for household insurance.
Using a proprietary model developed three years ago, the data mining team at Aegon Hungary determined that around 90% of its policyholders were 'reliable', meaning that they never submit a claim, or their claims are always error-free.
Automated claims settlement
Based on this knowledge, in co-operation with customer relationship management (CRM), claims and product development teams, Data Miners developed a model which made it possible to increase automated claims settlement from 20% to 40%.
The previous level of automated claims settlement (20%) was based on segments. The advantage of the new method is that it analyses customer behavior at contract level.
"Concretely, this means we will no longer send an assessor to investigate the claim further, but only focus on the payout," explains Dániel Szabó, head of the data mining team and a member of the Aegon Analytical Academy.
Less administration for customers
For most people, preparing and submitting a claim to an insurance company is an extra administrative burden in their already busy lives. By doubling the amount of claims that can be processed automatically, customer satisfaction is expected to soar.
"It's going to be a great customer experience," says Dániel. "It means customers won't have to wait for an assessor anymore, don't have to negotiate anything, and don't have to wait for their money."
Reduced risk, lower premiums
Increased automation will also benefit shareholders. Non-life insurance accounts for 60-63% of Aegon Hungary's profits. Greater automation in claims handling lowers the company's claims ratio, an important indicator of profitability for an insurance company. (The claims ratio is the percentage of claims incurred of total premium income.) And lower claims also lead to lower premiums for all customers.
Not all the claims of low-risk customers can be automated, however. Assessors will still get involved, depending on certain factors, such as the type of house a policyholder lives in, the type of claim they are making, and how high the claim is.
The team took the project a step further by developing an algorithm that can predict whether a new customer will make a claim or cancel their policy within the first six months of purchasing a non-life policy.
The first six months are a key indicator: claims submitted after this period are generally considered low risk. As a result, these can also be processed automatically, leading to quicker claim handling, and happier customers.
The model can be used for making the procedure of underwriting simpler and more accurate. The application of this model is currently under discussion with other relevant departments within the company.