The life insurance business is primarily about managing the following three aspects efficiently and effectively so as to achieve long-term profitability and return-on-investment objectives:
- Mortality risks
- Investment risks
- Operational expenses
The success of a life insurer depends on how well it manages mortality risk and operational expenses in order to keep costs in line with pricing assumptions, and on how it can maximise return-on-investment on its capital and assets. These business needs pose competing demands and irreconcilable constraints in front of corporate decision makers.
Achieving a balance among all three at the same time is a challenge and in this article, we will look into these challenges; in particular, those around the underwriting (UW) process faced by insurers, followed by illustrating how predictive analytics can help mitigate these challenges.
Challenges and role of analytics in life insurance
Underwriting is a gateway to protect life insurance companies from acquiring non-profitable businesses by appropriate risk selection and grouping of homogeneous risks. It acts like a detective for unfavourable proposals and is one of the prominent areas which directly impacts mortality experiences. It should be one of the areas insurers apply analytics as early as possible.
Enhancing the UW process by applying smart tools and technologies can result in favourable mortality experience and reduce expenses.
Let’s look into the generic UW process in Figure 1.
Real-time insights
Currently, UW is carried out in a very traditional way except for a few special instances. It constrains life insurers from optimising the value for their customers and for their businesses. It is now possible to overcome these constraints by enhancing the UW process with real-time insights from the broader UW horizon with the help of analytics. For example, insights of mortality experience by medical labs, geographic region, specific trend of diseases by geographic region due to extreme weather factors of climate change, crime etc.
Another aspect is to leverage the persistency insights by advising new prospects at the beginning to opt for favourable factors which could increase the persistency for their policy. Leveraging on production UW data to assess the consistency in UW decisions could also be explored. Above all, insurers need to look into what kind of customer experience they are offering now and how can they shorten the “time” and “space” element with a better customer experience in the overall process. It is proven that the companies who tackle these challenges are better positioned in the market among their peers.
Although they are not in the life insurance business, P&C insurers such as Esurance and Kroodle are notable pioneers who leverage technologies to provide an easy engagement process and better customer experience.
Intelligent underwriting
Other companies such as FitSense work as data enablers to make a variety of data available to the UW process. In fact, there are also a number of very sophisticated products in the automatic underwriting space such as Velogica, which claims to be very successful in their own right. All these examples demonstrate that intelligent UW is a real possibility.
One of the obvious constraints for life insurers – which may be one of the reasons holding them back from their next move – is to strike a balance between the cost of adopting these measures and the reward in getting it out within a stipulated time horizon. So what is the way forward?
Intelligent UW should be the part of the overall advanced analytics strategy at the enterprise level, which should ideally evolve with time.
- Analytics should be a long-term plan and acted upon by charting a roadmap.
- Organisations can begin with a small team for Analytics COE to lay down the foundation initially. This can help them optimise their existing processes without incurring substantial cost and also help them create awareness for analytics and a culture of making data-driven decisions. The Analytics COE can then do an enterprise-wide process mapping, defining metrics, setting priorities and gearing for the next phase.
- In the beginning, focus should be on the functions and metrics which start delivering value within a shorter time period, say weekly, fortnightly or monthly. We can call it dynamic metrics, metrics which can instantly improve if we take some remedial action at business front. It could be a subset of the long term plans, and as it progresses, brings credibility, authenticity and awareness among the staff for analytics. Each accomplished component would keep integrating to the long-term map of the overall analytics framework, and one day, the organisation will have a mature Analytics COE.
Predictive underwriting
If we divide the whole UW process into three steps, then the first step would be validating certain aspects straightway such as insurable interest, age, income, sum assured, occupation, habits, etc. The second step could be an assessment of medical test reports, and the final step could be assigning the risk category and any loadings.
We can apply the following techniques to bring about some sophistication in the UW space.
Data capturing and validation
At the outset, the effort should be to capture correct and relevant data. Companies should look beyond the data captured via application forms and extract data from various other sources. In fact, companies should strive to go digital so that the data can be available and validated in real time.
Decision Tree
Decision tree can be applied for the knockout cases. It can also be applied to classify cases into some defined categories based on defined rules.
Classification Rules
Supervised Machine Learning such as Classification Rules can be used to auto-classify the risk. A training dataset could be prepared afresh, or it could be extracted from the production for this purpose.
This is not the exhaustive list of techniques; the eventual need can be better identified at the time of real implementation.
Reinsurers also have a stake in this exercise. If a particular insurance product is reinsured, then consultation with reinsurers around application of these analytics techniques is also important in order to avoid any contractual issues later on.
Value propositions
- Auto-UW with predictive capabilities could free up some UW resources for other critical processes, such as claim administrations. It will help insurers contain some of the expenses incurred at UW, and provide better control on risk selection, which in turn results into mortality profit.
- Currently, there is a lot of discretion attached to UW decisions which might not be accurate sometimes; for example, a follow-up or further medical test is generated where it might not be required, impacting medical costs. If a follow-up is required, predictive UW could be helpful in processing it faster and directly conveying the output to the concerned agent and the proposer, shortening the overall process loop. It will eventually be well-received by the proposer and agents.
- Predictive UW could boost business performance at times of peak volumes eg during the financial year end.
- Above all, the overall process is supported by data and requires less manual intervention, hence the underwriting results could be less biased. Furthermore, if insurers leverage on the production data as a benchmark for the UW, it could give them a sense of confidence. After all, it is a mirror of their deliberate selection of risks.
Conclusion
Early adoption of analytics is imperative for life insurers to achieve sound UW process going forward. Today’s investments will help insurers adapt to the disruptions caused by newer technologies and tools, such as blockchain, genetic uw, cognitive computation, wellness devices, chat-bots, robo-advisor etc.
The overall analytics roadmap and capabilities have to be future-proof and ready to capture the market opportunities emerging due to the rapid pace of technological developments. A
Mr Kedarnath Jha is a Senior Business Analyst at DXC Technology, and a Fellow of the Life Management Institute. He is also an aspiring Data Scientist.