The need to make predictions in the life insurance space is nothing new – predicting future mortality trends is the very basis of life insurance risk assessment, after all.
Yet today this practice of predicting has taken on increased complexity, and data science may hold the key to unlocking what the future has in store.
Several factors have contributed to the emergence of data science in life insurance:
Any successful data science program requires three basic but essential components:
It all starts with the data itself, the raw material for data science. Information must be derived from reliable sources and provide enough quality and quantity of data with sufficient breadth and depth to enable meaningful analysis. Equally important, all data must be acquired ethically and in accordance with all applicable laws and regulations. Advances in applied data should be pursued with a heightened awareness of potential scrutiny and a sensitivity to privacy concerns.
While the amount and range of data available – prescription histories, driving records, activity levels (via wearable fitness trackers), etc. – continues to grow at an accelerated pace, insurers are discovering that simply having greater access to data doesn’t automatically lead to greater predictive power. That requires the knowledge and skills to understand and clean the data and then to convert it into actionable insights – data science, in other words.
Data scientists combine domain knowledge of math/statistics, computer science, and relevant subject matter (life insurance in our case) to develop predictive models that learn from historical data to predict future outcomes. They can discover complex patterns in historical datasets, and then apply those patterns to new data to make predictions about what will happen next. For insurers, such predictions can provide new opportunities for innovation and growth.
Yet data science is of little value to insurers on its own – it must be integrated into insurance processes and pursued in support of shared business goals if it is to fuel progress. Data scientists need to collaborate with teams throughout the company, leveraging the experience and expertise of product developers, underwriters, actuaries, claims specialists, and others to identify and implement practical data solutions.
This collaboration extends outside the insurance space as well. As companies from various related industries seek to provide healthcare services using digital technologies, data scientists within insurance companies must seek working partnerships with other sector experts – professionals in industries ranging from healthcare providers to tech startups – to develop new life insurance products.
The following examples provide brief snapshots of successful applications of data science.
Having already helped develop a successful approach for applying prescription history data to mortality risk selection, RGA’s data solutions team was tasked with developing an improved mortality index to make prescription data easier to use and interpret for underwriting. The new index integrates a range of prescription data – including severity, frequency, recency, and doctor specialty – to generate predictive risk scores. By linking this data to mortality experience, these risk scores help facilitate accelerated underwriting and, when combined with other underwriting information, have proven to contribute to more comprehensive and accurate applicant risk profiles.
An insurance company with a sizable cancer product customer base sought to cross-sell a new combined life and cancer product to in-force policyholders through a significantly simplified underwriting process. Using client-provided data such as individual policyholder demographic information, policy attributes, and claims data, a model was developed with RGA's support to identify the best risks among the in-force customer base for cross-sell opportunities, fast policy issue, and improved claims experience.
An insurer was seeking a consistent and effective fraud detection process for incoming claims that optimized limited investigation resources. RGA worked with the insurer and used experience data of already investigated claims with known results and combined it with customer demographic and policy information. The result is a predictive model that eliminates the need to investigate the best 25% of claims and triggers more vigorous investigation of the worst 25%. The model also provides greater insights into the driving factors of fraud cases that can be incorporated into pricing bases.
Data science has fundamentally changed the insurance industry and will continue to do so. Now is the time to identify ways to access relevant, reliable data and to develop the expertise to generate insights from that data. As the case studies above demonstrate, the potential applications of data science span the entire insurance process, from underwriting through claims. All functions within an insurance company should be exploring ways to leverage data science to distinguish their company from the competition. Actuaries in particular should seek to keep up with the latest developments in data science to broaden their perspective and expand the breadth and depth of their expertise – for the advancement of both their company and their career.
In an article in Harvard Business Review in 2012, “data scientist” was described as the “Sexiest Job of the 21st Century.” Since then, demand for data scientists has skyrocketed in many parts of the world, and that certainly includes Japan. Data is the currency of our modern world, and the insurance industry needs to embrace this new era.