Data science use cases in Life Insurance
Data science is a new phenomenon in the tech world today. With increase of computing power, it has changed the world how we live today. It is difficult to live without your phone with which you spend almost 20% of your time every day. Lots and lots of companies are using data science to increase their sales, for customer analytics, for campaign management, for identifying their best customers etc. When you like a post on Facebook or do a quick search on Google or get a sales call to buy a credit card or get recommendations of products when you buy something on Amazon page, a hell lot of things happen on the click of mouse or when you touch your mobile’s screen. It’s a brave new world when you are the end user and even for data scientists who do that magic behind that PC screen or underneath your mobile screen.
Data-driven analytics startups in insurance (e.g., Pilotbird) are leading the way in adopting publicly available data (e.g., social, census) to more accurately price life insurance risk, segment clients and detect fraud. Life Insurance companies are not far behind. Infect they were the main users of data science. Actuaries use prediction models for mortality and pricing the insurance products. When it comes to statistical packages, R is the favorite language for Actuaries. There are many R packages which are used in insurance like chainladder, cplm, lossDev, actuar, favir, mandate and lifecontingencies.
In India, there are more than 24 life insurance companies. Life insurance companies generate millions of rows of data every month. This data could be in the form of renewal premium, new business premium, accounting, leads, alterations and much more. Indian insurance giant LIC of India sells more than 20 million policies every year and is currently serving more than 310 million policies. So you can imagine the humongous amount of data generated by insurance companies in India and world over. This big data presents a great amount of opportunity and challenges to data scientists the world over.
Look at the premium income earned by Indian insurers below. You can imagine the enormous amount of data being generated every day by these companies.
But how life insurance companies can use this big data in an effective way as they are sitting on huge gold mine in forms of data. How life insurance can companies provide their customers the best service? How life insurance companies can generate leads for its sell force from existing data as it is always cheaper to sell a product to an existing customer then a new customer? What could be the use cases for life insurance companies to use data science?
I will show you some good use cases for data science in Life Insurance. This list is not exhaustive. If you are a Guru of life insurance then you can think of a lot more.
Policy Lapsation
Policy lapsation is the biggest pain area for life insurers. Almost 50 % of policies lapse in India before they complete 5 years. With the help of this model, customers can be contacted in advance who have high chances of lapsing their policies. Controlling lapsation will lead to higher premium income for the insurer, increasing commission income for the sales force, customer satisfaction and also a valuable risk cover for the customers.
Customer attrition
This data science model predicts the attrition of customers. Attrition could be in the form of surrender or lapsation of policies. An early surrender leads to loss of premium income for the insurer and also loss of risk cover for the customer. Surrendering a policy early involves lots of deduction of charges. The customer never gets back what he paid for. The model looks for the pattern in the data. The insurer also comes to know which of its products are not popular and can redesign the products as per customer’s requirement.
Product affinity
Product affinity studies behaviors of the customers so that they can be cross sold with new products. This data science model helps in targeting the customer with an alternative plan. For example a customer having a Ulip policy could be targeted for a traditional plan. A customer after age 40 can be targeted for a pension plan. The insurer can study the data of the customers visiting its web site. Some of them could be existing customers and some of them would be new customers. Looking at the pattern of the data, customers can be targeted with alternative products. You might have seen same thing when you go to Amazon site. In the bottom of the screen it recommends the products which you can buy basing upon your buying behavior.
Agent’s attrition
Life insurance agents do most of the business for life insurers. They are also called Feet on Street. Insurers spend considerable amount of money and time to train and hire insurance agents. If these agents leave in between then the cost incurred is never recovered. LIC of India has more than One million agents as a part of its sales force.
This data science model predicts the attrition of an agent basing upon Annualized premium of polices sold by him, no of policies sold by him every year, lapse policies, duration with the company etc. With the attrition of agent, chances are high that policies sold by that agent would lapse in future.
By looking at the behavior of agents, insurers can take suitable measures to retain them.
Renewal premium forecasting
In 2017–2018, Indian insurers collected almost INR 265000 cr Renewal premium.
This data science model forecasts the renewal premium to be collected on existing policies on the books of life insurers. This helps in setting the targets for the sales team and helps in maintaining persistency of the policies.
Customer Life Time Value
This data science model helps in determining life time value of a customer. The model determines the value of current policies owned by the customer and expected tenure of current policies. It also takes into consideration the new policies customer can buy and forecasting of premium earned on these policies. The lapsation model, repurchase model and forecasting model are also taken into consideration. The model also takes into consideration fixed cost associated with each policy. The model can later be used in cross sell, up sell and targeting the customers across different marketing campaigns.
Agent Life Time Value
This data science model helps in calculating the life time value of an insurance agent basing upon business done by him so far and expected business he can generate for the company in future.
Most of the policies in insurance companies are sold by an insurance agent. He is the link between the insurer and the customer. This model takes into consideration number of policies sold by the agent, first year premium income earned, likely renewal premium to be earned on the policies, commission to be paid and fixed cost involved for retaining the agent.
Once the ALTV is determined, insurance companies can take extras measures to retain top performers by giving them extra incentives.
Orphan customer identification model
An insurance agent is an important link between the customer and insurance company. If agent leaves the insurance company, there is no one to serve the customers of that agent which leads to lapsation of policies sold by that agent and in turn leads to loss of revenue for the company. This model helps in identification of orphan customers.
Companies can identify such customers and can assign the policies to the active agents thus controlling lapsation and retaining customers.
Customer Segmentation
Customer segmentation is the process of putting customers in same bucket based on similarities. In life insurance, there are many variables like Face Amount, Frequency of Premium Payment, Risk category, Product Types, City of Residence, Occupation, Annualized Premium, Posh Locality Resident etc on the basis of which customer segmentation can be done. This is one of the most important data science use case in life insurance.
This data science model segments the customer basing upon different criteria as mentioned above. It puts the customer in different buckets which in turn helps in targeting the customers for different marketing campaigns. Insurance companies can also design their products as per customer segmentation i.e. right product for right customer.
Different techniques like Factor Segmentation, K-means clustering, Two step cluster analysis, Hierarchical clustering and Latent Class Cluster Analysis can be used in customer segmentation.
Claims Prediction
Claims management is an import activity in life insurance. Claims can be of different types like death claim, disability claim, critical illness claim etc. Companies can build a predictive claims model by looking at the types of claims paid so far. Risk categories from an underwriting risk model should also be used to build claims prediction model. Once a claims predictive model is built, companies can drill down the data to find the important pattern like from which region the claims are coming, which agents policies’ incur maximum claims and portfolio of products where maximum claims have come.
Underwriting Risk model
Underwriting is the first entry point where risk assessment of proposal is done. Depending upon the type of risk, the life to be insured possesses, the insurance companies decide to grant the risk coverage to that person. Coverages can be given without charging any types of extras or if the person possesses any type of risk like occupational risk, health risk etc., he could be charged extra amount to cover that risk. Insurance companies can put different insured in different types of risk.
An underwriting model can be built which will predict risk on new policies which are being sold daily. This will help insurance companies in underwriting of the proposals quickly thus saving the cost of medicals and health tests and also savings in terms of future claims.
Conclusion
We have seen that there could be different use cases of data science in life insurance. And this list is not exhaustive. More use cases can be added to this list. Life insurance companies have been using predictive modeling since long time and will continue to do so. Pricing of insurance products is done based upon predictive modeling. With advancement in data science, insurance companies can use new algorithms to price their products thus increasing the profitability and retain their sales force and customers.
Sources
Contact me at harnagpal@gmail.com for any query.
