How insurers can improve data privacy in ecosystems

Introduction

Insurers have traditionally used historical data to predict future risks.

With the proliferation of digital technology, they have significant potential to capture data at every customer touchpoint, including third-party sources. The real challenge is the integration of data insights contributed by various players into the insurance ecosystem, as data sharing presents a dilemma. While data can allow insurers to offer personalized products, customers are worried about its safety and usage. And privacy laws such as the EU’s General Data Protection Regulation (GDPR) validate customer concerns. In this scenario, insurers are looking for ways to counter data privacy fears by embracing emerging technologies such as privacy enhancing computation (PEC). The technology leverages cryptographic encryptions, data masking, and artificial intelligence (AI) and machine learning (ML) algorithms to ensure no sensitive information is shared. Yet, it allows stakeholders to extract actionable insights such as customer preferences and habits from the data. These can be integrated for cross-selling, repeat sales, and customer retention. Insurers can use privacy enhancing computation in a number of ways. 
 

Telematics

Usually, drivers buying insurance make a trade-off between data shared and premium discounts received from insurers.

Often, it is not worthwhile for them to share data since the scoring on driving skills and other rules are stringent. Hence, eliminating privacy concerns can improve usage-based insurance (UBI). This can be achieved by generating synthetic data that helps design risk models. Additionally, there is always support from data providers in the ecosystem in the form of satellite data, demographic data, and so on. With a combination of these data sets, insurers can artificially generate synthetic data. Data collated from drivers’ claims and associated classical insurance risk variables and telematics-related variables can be factored in to create synthetic data. It replicates real data and identifies individual behavior patterns or statistical properties in the sequence of events. These can support AI and ML models during simulations to generate insights. 
 

Home insurance

Home insurers can leverage data for underwriting.

However, the Central Consumer Protection Authority (CCPA) and Health Insurance Portability and Accountability Act of 1996 (HIPAA) regulations present challenges for data scientists and actuaries. They restrict the use of crucial data such as home addresses. This has been hindering modelers in creating optimal assessment of risk and offering better pricing. Insurance organizations can create synthetic geolocation data and home addresses and use them in simulations for floods, storms, fires, earthquakes, and so on. The pricing model based on synthetic data is as accurate as that based on real data. It offers customers better prices. Synthetic data can also be combined with some real data. This is true not just for addresses but also for replicating images of homes and accurately identifying features of a property such as condition of roof, existence of pool, and so on. Data from various market players in the ecosystem supports this method.
 

Electronic health records

Electronic health records reduce the need for invasive medical examinations during underwriting.

Several market players contribute toward making this data ready for consumption. They could be service providers at the point of care, laboratories, and data aggregators offering health record platforms. They also include tech firms offering analytical insights and insurance organizations finally consuming the data for underwriting and optimal pricing. With active access, contributions, and sharing, all participants can boost predictive performance. 

To mitigate privacy issues, insurers can apply privacy enhancing technologies such as multiparty homomorphic encryptions and differential privacy. These technologies are evolving quickly and eliminate or restrict the need for sharing personally identifiable information. Normally, these data sets are shared to centralized pools during data aggregation, and analytics are derived from the pools.
 

Conclusion

Insurance firms and large data providers must adopt privacy enhancing computation.

This will facilitate the provision of services by many players in the ecosystem. It will lead to the creation of a large reservoir of data that is continuously updated, forming a rich source of analytics and insights. Insurers can realize several benefits such as accurate yet non-invasive medical underwriting, better pricing, and enhanced customer experience. All this will be made possible without the need for customers to share their personal or reconstructed data, either during aggregation or resharing. Privacy enhancing computation will bring business agility not just to orchestrators or insurance organizations, but also to other players in the ecosystem.