Free Preview

This is a preview of some of our exclusive, member only content. If you enjoy this article, please consider becoming a member.

This is the second in a series of articles by Valen Analytics looking at the hurdles that insurers must overcome to effectively implement and gain value from data analytics programs.

After deciding to implement predictive analytics, insurance leadership will likely need to address the operational issues of bringing an analytics program from strategy to production. One such hurdle for the C-suite is encouraging underwriters to adopt new data solutions to improve their understanding of risks across a book of business.

Executive Summary

The C-suite has some challenges ahead in finding ways to encourage underwriters to adopt predictive analytics in a way that helps them better understand risks. Kristin Marr, president of Valen Analytics, writes here about a number of things executives can do to boost underwriter buy-in of the technology.

Barriers such as institutional culture or fear of being replaced can lead underwriters to work around systems, or even avoid them altogether. While largely inaccurate beliefs may manifest this behavior, these concerns can be overcome by executives that take the time to target their analytics initiatives on measurable performance metrics and include all stakeholders in each step of the implementation process.

1. Early Inclusion

Editor’s Note: More recently, Willis Towers Watson reports that 80 percent of the U.S. carriers and 79 percent of Canadian carriers participating in the 2017 Advanced Analytics survey indicate they use advanced analytics in some form or fashion, whether it be underwriting/risk selection and/or rating/pricing, or claims, or marketing or other applications.
A majority of insurers (60 percent) now use predictive analytics in their everyday processes, according to a 2016 study by Willis Towers Watson. This is in-line with Valen’s 2018 Outlook Report that found 61 percent of insurers use analytics for underwriting based on a three-year average. However, not every underwriter has bought into new approaches, with more than half of front-line employees showing resistance to new technologies.

There are two steps that insurers can take early in the process of rolling out analytics programs.

The first step to achieving full buy-in is to identify underwriting team members with a robust knowledge of the book of business, but also interested in trying something new. Institutional knowledge is critical for understanding variances between the suggested pricing of a model, and the historical way an insurer has priced a policy. Studies show that insurers using analytics outperform those who are not, but a combination of data and individual knowledge yield the strongest results.

Once a team is in place for the pilot program, it is critical to communicate and celebrate successes that result from predictive analytics programs. If an insurer can improve the bottom line and grow top-line revenue by winning the types of business that fall into its risk appetite, this information should be disseminated across the team and organization.

Insurers must remember that their underwriting staff needs to undertake a mindset shift to fully incorporate predictive models in their day-to-day workflow.
The second component of early buy-in focuses on creating effective workflows. This goes beyond offering underwriters a new spreadsheet to incorporating predictive analytics solutions into an underwriter’s day-to-day work environment.

It’s important for the developers and IT team building analytics solutions to understand the steps an underwriter takes before creating a policy. To do so, IT can perform interviews or even serve as “daily shadows” to underwriters, which allows them to see common approaches to understand how and when exceptions to typical workflow may occur. This information should be used to build the data analytics process or solution to lower the barrier to entry for underwriters to incorporate new data.

2. Provide Line of Sight

Aligning an organization around analytics initiatives means effectively communicating the goals for a program at the onset and explaining how a model aligns with meeting those goals. For example, if an insurer wishes to grow their workers compensation business into new states, predictive analytics can help identify adequate pricing for policies despite never writing business in the location before. When the necessary historical data is not available for underwriters, analytics is a powerful tool giving them greater confidence that they can hit their target goals.

Another key consideration is conveying how models can empower underwriters. Some underwriting teams are restricted in the amount of credit they are permitted to offer, whereas using a model will increase their flexibility. The use of analytics is meant to improve the work environment for underwriters, not replace them.

3. Build Feedback Loops

This article is the second part of a series of articles by Valen Analytics. In Part 1, Marr offers some warning signs of predictive model bias and methods to overcome it.
One of the toughest barriers to overcome during implementation is knowing what information to track and being able to address feedback from all stakeholders. Insurers can gather policy and model usage data from underwriters in the early stages of an analytics program. It’s important for insurers to learn each instance where underwriters believe that the output from the model fails to align with its stated goal. This insight is critical because it allows insurers to identify policies written that may fall outside their risk appetite. Also, if enough underwriter concerns accumulate about certain types of business, it is possible that there is bias in the model.

The use of analytics is meant to improve the work environment for underwriters, not replace them.
Equally important for insurers is the model usage data. This metric identifies how often a model is being consulted and used by individual underwriters. Management will have to walk a fine line between model usage and model overreliance. It must identify the gap between underwriters who don’t use it at all, which eliminates its effectiveness, and those who rely on it entirely, limiting the value of underwriting experience and institutional knowledge. AIG has highlighted the importance of this in the past by referring to underwriters who consistently price at the model as “flat liners”.

4. Forget About the Machines

Aside from finding the right test groups and giving them the ability to provide the necessary feedback, insurers must remember that their underwriting staff needs to undertake a mindset shift to fully incorporate predictive models in their day-to-day workflow. It can be tough to implement new approaches for individuals who have been working the same way for years, both in insurance and in most other industries. However, through a phased rollout that celebrates successes and addresses concerns, insurers can secure acceptance across their underwriter base. This will ultimately lead to increased growth, lower claims, and a book of business that more accurately aligns with an insurer’s risk appetite.