Part 2: Predictive Analytics
Learn about the SOA’s efforts on predictive analytics involving education, employers and professional developmentMAY 2019
Leaders from the Society of Actuaries (SOA) gathered at the Schaumburg, Illinois, headquarters to reflect on the past year and to share their perspectives on the future of the organization and the actuarial profession. These discussions were with 2017–2018 SOA President Mike Lombardi, FSA, CERA, FCIA, MAAA; 2018–2019 SOA President James M. Glickman, FSA, MAAA, CLU; 2019–2020 SOA President Andrew D. Rallis, FSA, MAAA; and SOA Executive Director Gregory W. Heidrich.
We previously shared their discussion of the SOA 2017–2021 Strategic Plan and key activities from 2018. In this second part of the series, we share their discussions on predictive analytics.
Q: Predictive analytics is “much talked about” everywhere these days. How did it impact the SOA in 2018?
Heidrich: The SOA Board of Directors has clearly identified predictive analytics as the big issue to address. We have a three-pillar strategy that comprises the fundamentals of what we’re doing and how we’re going to do it.
The first pillar is our decision to add significant predictive analytics content to the SOA’s prequalification education. With this change, we can assure employers that our future members will have demonstrated a strong grounding in the fundamental concepts, techniques and tools of predictive analytics. We’re making a promise that actuaries who come through our exam path will have predictive analytics training.
We are particularly pleased with the creation of the new predictive analytics assessment, which requires a candidate to demonstrate that they can look at a data set and figure out what it might be able to tell them to solve a real problem. They then work with statistical software to convert that into analysis and end by writing a report on their analysis and findings. They do all of this under the pressure of a high-stakes, professional exam. While candidates do all those kinds of things in their work life, they now can show it on an exam. We’re exploring an expansion of this testing capability to other exams.
Another pillar of activity is focused on helping employers understand and know that actuaries can do this predictive analytics work. This marketing pillar started in the health field and more recently moved to the life insurance field. Its purpose is to demonstrate to employers across a range of fields that actuaries have these skills and should be the go-to source for predictive analytics and problem-solving.
As part of this effort, we encourage members who are interested to participate in Kaggle competitions—modeling and machine learning challenges posed to entrants from any discipline. We view the SOA Kaggle Involvement Program as a way for our members to showcase their skills. Our members have been successful with these data science projects, and some have reached the Kaggle Master Level, which is difficult. This demonstrates a point Jim made earlier in our conversation that if you turn smart people like our members loose on some of these problems, they have a great deal to contribute. We want to encourage our members to take on problems that may be outside of their comfort area, and we also want to demonstrate to the people who watch these competitions what actuaries can do and how good actuaries are.
Another pillar is professional development for our existing membership. We have many members who need to be and want to be familiar with predictive analytics concepts for their roles. They may do the work themselves, or they may supervise it. They may be consumers of the work and need to understand it. We’ve been adding more predictive analytics content to our professional development offerings, and we created the SOA Predictive Analytics Certificate program for actuaries to demonstrate these skills in the market.
Our certificate pilot in 2017 with 30 members was successful, and this past year we ran three sittings with around 50 people each time. We’re planning to do more sittings for the certificate this year. Members are attracted to it; they’re looking to get this knowledge. Keep in mind it’s not an easy course. It’s a serious commitment of both money and time for participants, and we’re excited to see this commitment.
Lombardi: Yes, our Predictive Analytics Certificate program has been quite successful and has become a standard SOA offering. It allows actuaries who have already finished their exams to get up to speed on the topic. This five-month program is an additional learning opportunity for seasoned actuaries interested in predictive analytics.
Heidrich: We also have the annual SOA Predictive Analytics Symposium that brings together more than 200 actuaries, and we recently co-sponsored a predictive analytics seminar in Toronto with the Canadian Institute of Actuaries (CIA) and the Casualty Actuarial Society (CAS). All of this is exciting and are examples of what we’re doing to prepare our members for working with predictive analytics.
Lombardi: If you look at the predictive analytics techniques, there’s certainly a more technical side that appeals to many, but I see a dual path where some actuaries will choose the more technical route and others will do the more generalist business acumen type route. We must be broad enough to excel at both of those areas and make sure that people’s careers are satisfied either way.
Rallis: It’s important context to know how we got here. We’re entering an era where big data sources are important. The role of predictive analytics is around the analysis of very large data sets, which are becoming available because of biometrics, social media or genomic data—things we never had before. Actuaries have been trained in many aspects of data analysis, but it was usually around being able to make extrapolations from small data sets. Statistics was developed centuries ago to extract maximum information from limited amounts of data. Now we have more data than ever, and there are different techniques needed to weed out what is truly information from what is noise. It’s different than what actuaries have traditionally been trained for in statistics, and it’s necessary going forward.
Heidrich: Andy, I’m curious about your work at MetLife and data analytics. How is all this data leading to changes?
Rallis: These changes give us the opportunity to enter the areas of pricing phases, underwriting and claims processing. It’s helping us better differentiate between policyholders. You’re able to correlate information that may never have been available before regarding claims, and you can draw unique inferences. It leads to different product designs, pricing structures and underwriting classes.
Glickman: Actuaries are trained in a way that helps them question things coming out of numbers that don’t look right where you might otherwise come to a pure analytic answer. They’ll also learn how to find new and better ways to use the data. Being able to figure out new and better ways to apply that data, to subdivide it, to combine it—all of this leads to a better end product, and that’s part of the analytical work that actuaries do, especially in the product development area.