This article originally appeared in the CIA (e)Bulletin.
by Jean-Yves Rioux, FCIA
Predictive analytics offers the tools needed to implement a wide variety of valuable applications in the life and health insurance industry. How are Canadian life insurance organizations taking advantage of these tools?
A new study commissioned by the Society of Actuaries and the Canadian Institute of Actuaries has found that most Canadian life insurers face key challenges related to data quality, centralization, and procedures that hold them back from unleashing the full potential of predictive analytics.
| Predictive analytics: applying statistical techniques to calibrate a model using historical data to make predictions about future or otherwise unknowable events. |
The Use of Predictive Analytics in the Canadian Life Insurance Industry study, conducted by Deloitte, polled 15 Canadian life insurance organizations that were a mix of direct writers, reinsurers, and bank-owned insurance subsidiaries. The study is the first predictive analytics survey in the actuarial community exclusively focused on the life and health insurance sector in Canada.
The study revealed that the most valued applications were those relating to simplified underwriting, fraud detection, targeted marketing, and claims management optimization, among others.
Having a comprehensive data repository with clean, complete, and accurate data is critical for deriving full value from any predictive analytics program or initiative. However, just 27 percent of the survey respondents indicated they have some data centralized at a single point in the organization, indicating a significant gap for the majority of the industry.
What’s more, many participants, especially those from mid-sized and small companies, also do not view their data as particularly complete or accurate. Respondents from large organizations ranked the completeness and accuracy of their data at 4.0 out of 5. Mid-sized and small life insurers, however, scored their data at only 2.8 and 3.0 out of 5, respectively, again signalling room for improvement.
All of the respondents who are currently undertaking predictive analytics projects said they use third-party data from vendors and government agencies to supplement their own. Almost three-quarters (73 percent) use demographic data, 60 percent use geographic data, and more than half (53 percent) use claims and medical data and credit data provided by outside data providers.
Canadian life insurers also appear keen to use new technologies, such as wearables and sensors, as a source of data for predictive analytics. One-fifth of respondents said they collect and use data this way, while almost half (47 percent) are currently investigating the possibility of using sensors and wearables such as fitness trackers or smart watches.
| Survey participants BMO Insurance Desjardins Financial Securities Empire Life Great-West Life Assurance Company Industrial Alliance ivari Manulife Munich Re Optimum Re Partner Re RBC Insurance RGA SCOR SSQ Sun Life Financial |
Policy priorities
The study also polled life insurers about their governance policies concerning data used for predictive analytics. Given the private and sensitive nature of the information they handle, all respondents indicated that having data privacy and data security policies in place is crucial. There is significant business, legal, regulatory, and reputational risk associated with exposing customer information – or having it stolen or hacked – and it is clear the industry has made managing this risk a priority.
At the same time, there is less emphasis on governance policies that can assist with predictive modelling. For example, 60 percent of respondents said they have data accuracy and quality standards in place, but just 33 percent had policies or governance in place regarding data standardization.
The survey revealed that another key to success in deploying predictive analytics initiatives is dedicated leadership and support from top leadership. Overall, 60 percent of respondents said they have an executive data leader and a predictive analytics leader in their organization, and more large companies have such leaders in place than their mid-sized and small counterparts.
It’s perhaps unsurprising then, that large companies placed a higher priority (4.7 out of 5) on analytics than their mid-sized and small counterparts (3.1 and 3.0 out of 5, respectively). The companies that did not rate analytics as a top priority cited the existence of other project priorities (50 percent), the low quality of internal data (40 percent), the challenges in obtaining data from multiple internal data sources (40 percent), and the need to prove ROI (40 percent) among their reasons.
Analytics teams
The study also offered insights about the size and composition of Canadian life insurers’ analytics teams. The average number of full-time equivalent (FTE) roles working on analytics was 10, with large respondents boasting larger teams (25 FTE) than mid-sized (seven FTE) and small (two FTE) respondents.
On average, statisticians and analytics experts comprise about 40 percent of the analytics headcount at Canadian life organizations. Nearly 50 percent are business experts, including actuaries and non-actuaries alike. Data architects and engineers accounted for 8 percent, and computer scientists and IT staff made up 5 percent of the analytics team.
While this first study of the use of predictive analytics by the life and health insurance sector in Canada shows that most companies have some challenges to overcome before they can take advantage of the full potential of predictive analytics, there are promising signs – particularly in the larger companies – that they are moving in the right direction. If all life insurers can stay focused on the benefits, the short-term heightened effort will be worth it.
Read the full study. Listen to the podcast.
Jean-Yves Rioux, FCIA, is Chair of the Predictive Modelling Committee.
