When buying health insurance, the agent or insurer will ask for your age, race, marital status and habits like smoking, health conditions, etc. Only after that will the insurer tell you what premium you need to pay for the insurance coverage.
Now it turns out, part of the job of actuaries is to try to predict the cost to the company to insure you for the life of the policy.
As technology advances, so too are the options for insurers to create these models as part of the underwriting process.
Gartner defines predictive modelling as a commonly used statistical technique to predict future behaviour. Predictive modelling solutions work by analysing historical and current data and generating a model to help predict future outcomes.
Predictive modelling in insurance
Kenneth Koh, global principal and director of insurance and financial services at SAS, says predictive modelling is collecting data from internal and external sources to better understand and predict the behaviour of customers.
Predictive modelling can help insurance providers reduce issues and underwriting expenses, better manage customer relationships and claims, as well as increase their sales and profitability.
He argues that the pandemic has shown to insurance providers the importance of predictive modelling, by allowing them to anticipate changes, and define rate changes and new products more efficiently. This is the essence of predictive modelling in the insurance role.
While the earliest use cases of predictive modelling were around customer insights and marketing, Koh says the same tools are now being applied to underwriting and claims, customer lifecycle management, sales and distribution, new business underwriting, and operational risk management.
Insurance industry challenges
After 2020, Koh cites three challenges currently impacting the insurance industry in Asia:
- Revenue growth and profitability management
- Financial crime and fraud
- Regulatory compliance.
In each of the above, Koh believes the fundamental challenge to insurers is how to make intelligent decisions. He is confident analytics can help resolve some of these business challenges.
“Having good data sets, a single view (single source of truth) of the customer, applying some form of AI/ML capabilities to build some of these models will help address some of the business challenges insurers face,” he opined.
Role of CIO in predictive modelling
Koh acknowledges that the traditional role of the CIO and IT team relegated to the technology infrastructure and the platform, and the data as well that resides in their data warehouse. He stressed, however, that where it involves introducing predictive modelling tools and techniques, IT is responsible for integrating and operationalising these into the enterprise technology infrastructure.
On-prem or cloud?
Is it better to leave the predictive modelling on-prem or move it to the cloud? Koh acknowledges that the cloud offers some cost-savings by taking away the cost of hardware procurement. The cloud provides some form of high-performance analytics infrastructure as well to support it.
When deployed on-premise, the insurance company will own that particular software and therefore they have to maintain it. However, at the end of the day, the on-prem or cloud decision is dependent on the company’s IT strategy.
He concedes the growing trend towards the cloud to extend the capability so that all business users can leverage the technology.
Objections to use of predictive modelling tools
According to Koh, having the right skills in-house is often used as an early excuse to reject the use of predictive analysis. He cited the example of a customer that initially thought of predictive analysis as the same as business intelligence reports.
Arguably the more realistic challenge for insurers considering predictive modelling is recognising where to start. Koh suggests identifying “low hanging fruits” like revenue growth, profitability management, and detecting fraud.
Criteria when considering predictive modelling vendors
For Koh, it is about technology, people and the process. He recommends taking an open integration approach when evaluating vendors’ solutions. He also suggests taking a partnership approach in the choice of vendor.
“It is important to work with a vendor that has experience in the industry. The deciding factor is to work with a vendor that understands my business. One that knows my pain points and can help me solve it using technology as an enabler.”
Kenneth Koh
However, he warns about the risk of short-term partnerships when it comes to technology to give the company time to understand how to scale this technology up to meet their business needs on an enterprise-wide level.
Final thoughts
Koh says insurance is moving towards personalisation. “Hyper personalisation is going to change the equation of trust between the insurance and the customers. Imagine being offered an insurance product and service that unique to you. We believe you may be interested in these services or offers we are providing through predictive modelling,” he concluded.
Click on the PodChat player and listen to Koh’s take on how data modelling will elevate the insurance business.
- Let’s start off with a definition: what is predictive modelling in the context of the insurance industry? What is the difference between predictive analytics and predictive modelling?
- In the context of insurance, what are say the top 3 challenges of insurers in Asia today 2021?
- Where does predictive modelling come in?
- At an insurance company, who decides what predictive modelling solution the company will use?
- Does it matter if it's an on-prem or cloud solution? What are the pros and cons of either direction?
- Can you cite the 3 most common misconceptions re predictive modelling?
- Most common objections against predicting modelling tools?
- There are many vendors out there with predictive modelling solutions. Without mentioning SAS’ offerings, what should the top 3 criteria (considerations) for deciding which solution to choose? Do we need to do a pilot or POC?
- In the process of adopting a predictive modelling solution, do you face issues around integrating with existing technologies as well as perhaps an insurer’s digital transformation initiative?
- How do you see predictive modelling adoption in the region in the coming years? What will accelerate or slow adoption?
