Photograph: Scott Lowden
Q: Why did you become an actuary? What attracted you to actuarial science?
A: I learned of the actuarial profession while I was doing graduate work in mathematical statistics at Boston University. After doing some research, I felt the actuarial profession would allow me to apply my statistical background to solve real-world problems. In my first position in pursuit of an actuarial career, I applied my knowledge of statistics to build a capital asset pricing model (CAP-M) that could be applied to mutual insurance companies. This was a very challenging theoretical exercise because mutual companies don’t have an observable measure of risk that is equivalent to Beta observable for publicly traded insurance companies. The model I built suffered some shortcomings and was eventually scrapped, but the project convinced me that I wanted a career as an actuary.
Q: What prompted you to choose a nontraditional career path?
A: During my tenure at John Hancock Mutual Life Insurance Company, I was always attracted to projects requiring statistical and data analytics skills. I enjoyed working with information technology (IT) professionals to help them interpret the actuarial content of actuarial projects so they could correctly program actuarial solutions. During this time, I learned a lot about IT architectures and protocols. I found a common language with IT professionals, and I learned to appreciate them for their talents as much as they appreciated my ability to translate actuarial project requirements into terms they could understand.
Q: Would you provide some work history and how it segued into your interest in predictive analytics?
A: At one point I held a position as Statistician and Branch Chief of Field Operations with the United States Department of Agriculture (USDA). It involved very little insurance work, but it was very fulfilling to be a part of the team building the predictive analytics engine for the USDA’s new Public Health Information System (PHIS). The PHIS is the food safety inspection system for the United States and is responsible for detecting the presence of residues and pathogens in the nation’s food supply to prevent foodborne illness outbreaks. Working on the PHIS was one of the most satisfying positions I have ever held. Before the PHIS, there was a significant lag in the reporting of violations at slaughterhouses and food processing establishments and in the notifying of appropriate agency personnel of anomalous lab findings resulting from statistical sampling. My group developed statistical sampling techniques to determine the amount and frequency of laboratory testing necessary to identify threats to food safety. Reporting and data surveillance are now done on a near real-time basis to detect anomalies in collected data and lab testing results. When anomalies are detected, alerts are sent immediately to the appropriate agency department within the USDA, trigging corrective actions.
My interest in predictive analytics grew stronger after my stint at the USDA. I went on to lead a team of predictive analysts at a Property & Casualty (P&C) insurer. Despite a lack of prior experience with P&C insurance, I was hired to lead a small group of predictive analysts to rebuild the company’s predictive models. There are natural synergies among the predictive modeling, underwriting and actuarial pricing areas of a company, and each has a vital role to play in building predictive models. The models I was hired to replace were not built using an interdisciplinary approach and suffered from a number of shortcomings. As a result, they were not trusted by key areas of the organization. The new models were trusted and understood by key areas of the company, because it was a company effort to develop, vet and test them.
Q: With regard to predictive analytics, what skills positioned you for work in this area?
A: Without question, you need strong theoretical statistics skills. A modeler needs to be able to distinguish between statistical noise and true statistical signal in underlying data. Speaking of data, a modeler needs strong data modeling skills, which requires an ability to understand when statistical results fail to replicate significant business relationships.
Being able to work with an interdisciplinary team of professionals also is a must, and this requires strong communication, interpersonal and interviewing skills. It is often the case that people do not know how what they know is helpful to others. Modelers with little background knowledge of a business can use developmental questioning techniques to determine whether important business relationships are patterned in modeling data and research causes when those patterns are not present. The requisite communication style needs to give people a sense of inclusion in building predictive models in order to maximize the utility of the business experience of other team members. In summary, to work in this area, a modeler needs strong statistical, data mining and modeling skills, developmental questioning skills and an inclusive communication style—because models don’t build themselves, people do!
Q: What skills do actuaries bring to predictive analytics that other professionals may not?
A: Actuaries are uniquely qualified to model complicated insurance concepts. They possess the requisite mathematical skills to build models that simulate insurance processes and reactions to economic conditions. The profession is moving toward measuring and assessing policyholder behaviors that adversely impact the quality of acquired business. Policyholder behavior is becoming more important for assumption-setting in life insurance and annuity lines of business. Actuaries have a lot to gain by observing the work of underwriters and agents in closing insurance sales. The sales process is a behavioral exercise that requires being able to identify what motivates individuals to buy an insurance product. Actuaries can become more “streetwise” by observing how insurance transactions are executed. The behaviors of both parties to the transaction, including those of agents, will help explain a number of the patterns in insurance data.
Insurers are applying principles from behaviorism to influence the transacting of insurance contracts. At a predictive analytics conference I attended last year, a life company discussed how it used behavioral concepts like social norming and choice architecture techniques in the design of its website. These techniques can nudge honest responses to underwriting questions and steer customers to products that best suit their needs. Health care exchanges have been using choice architecture principles since the adoption of the Affordable Care Act. The labeling of health plans as gold, silver or bronze has been shown to influence a consumer’s choice of a health plan. Consumers will often select a gold plan, because gold is perceived as more valuable than bronze, despite the fact that the gold plan may not align well with their needs or financial resources. Actuaries are going to need to become more astute in the application of behavioral principles in predictive modeling as well.
Q: What opportunities do you see in the life sector for predictive analytics?
A: P&C actuaries can leverage predictive modeling in underwriting, marketing, assumption-setting and financial modeling for their companies. Insurance scoring models build upon credit models developed by credit rating agencies like TransUnion, Experian, Equifax and Dun & Bradstreet for modeling in commercial and personal lines of insurance. While these agencies possess an abundance of credit transaction data and the skills to model it, they often lack familiarity with insurance data and regulatory issues necessary to build and gain regulatory approval for insurance scoring models. Insurance scores can inform the underwriting process, assist with tailoring marketing programs and increase the power of financial models.
I think underwriting in the life sector is going to move toward predictive models that couple lifestyle attribute data to medical underwriting data to assess the extent to which a potential insured makes healthy lifestyle choices. Some life companies are already collecting data on policyholders through wearable devices and rewarding their healthy behaviors with premium discounts, gift cards, travel rewards and much more to incentivize them to engage in behaviors that improve their state of health.
Technology-driven insurance products have been successful in luring younger generations to buy life insurance products. The industry desperately needs to appeal to younger generations for future revenue streams. This generation likes quick and easy app-driven solutions. An example of a company taking advantage of mobile app technology is Lemonade Insurance, considered by some to be the first U.S.-based “peer-to-peer” (P2P) P&C insurance company. The P2P model is characterized by, among other features, a digitization of the insurance process, quickly issuing policies and paying claims, low cost for coverage and the return of some portion of the premiums through a “giveback feature.” The life industry may be a ways off from the P2P model, but life insurers are definitely moving toward digitization to respond to consumer preferences.
Q: How is working in predictive analytics different from a more traditional actuarial position?
A: I am not confined to building predictive models for life insurance. My skills are transferrable to building predictive models in health and P&C insurance, banking, government and any other industry needing data-driven solutions to business problems. As a consultant, I most recently worked on a predictive analytics project for a medical supply company. There were anomalous patterns in its data that were leading management to poor decision-making regarding its sales force. I was able to distinguish “fake” patterns from “real” patterns and advise the company on the treatment of both in its decision-making and recommend IT architecture changes to improve the quality of its current and future data.
Q: How do you see the role of predictive analytics changing in the next five to 10 years? Where will actuaries fit into the equation?
A: It is already changing, and big data analytics is leading the way. Big data analytics is simply data reduction techniques consisting of descriptive analytics, predictive analytics and prescriptive analytics. It is not enough to classify and score risks using descriptive and predictive analytics; the field is moving in the direction of developing prescriptive analytics that iteratively improve predictive models in near real-time and indicate corrective actions to mitigate risks. A major goal of this effort is to turn unprofitable customers profitable and identify consumers likely to select against the insurer.
Other industries are already building prescriptive models. The insurance industry is not too far behind, but it is definitely in catch-up mode, in my opinion. It is important to follow what data scientists are doing in other fields to inform on the developmental efforts in the insurance industry. If the current activity to build data science shops by companies like MetLife and New York Life is any indication, predictive analytics is fast becoming integral to decision-making by senior management. Many may soon wonder how on earth the insurance industry ever lived without big data analytics.
Q: What advice do you have for people who may be interested in positions in predictive analytics?
A: Learn the business for which you want to build predictive models. It is really important to understand how the business is transacted from the point of sale and related distribution channels to how the business is coded in company systems. Understand who the “hunters” are and how they transact business, and who the “gatherers” are in the company and the controls that govern the work they perform. Learning these roles is vital to understanding company data and how it got that way. The hunters in an organization acquire the business, while the gatherers prepare and process data related to the business. The modelers are the scavengers who pick through the data and prepare it for consumption by decision-makers responsible for ensuring the future profitability and stability of the company.
Q: What are some of your best professional experiences/memories as an actuary that may inspire others to explore different actuarial paths?
A: Working for the USDA ranks high on my list of experiences actuaries should consider when exploring different industries where their skill sets are highly valued. Also, working in Washington, D.C., is pretty exciting if the political machine on Capitol Hill fascinates you. There is always a protest in D.C. in which you can take part. Seriously, though, I took pride in helping build a predictive analytics engine to monitor and protect the nation’s food supply from the invasion of threats from residues and pathogens. Banking also offers some exciting challenges for actuaries, and actuarial skills are highly respected by banking professionals. You will be surrounded by a lot of mathematical talent, and you will fit right in.
I encourage actuaries to teach actuarial classes for universities or local actuarial clubs. I taught in the actuarial science program at Boston University for seven years and won the outstanding teaching award in 2002. Boston University likes hiring working actuaries as adjunct professors because of all the actuarial work experience they can impart to students. Actuaries who teach, on the other hand, benefit by learning how to explain complicated actuarial concepts in simple terms. If successful, those acquired skills will transfer in communicating your ideas and the results of your work to senior management. When you communicate actuarial results in crystal-clear terms to senior management (many of whom will not be actuaries), you empower them to make better decisions and you become a trusted adviser, someone they call on again and again for their analytical needs. It is important to always meet people where they are in explaining technical actuarial results.
Q: What is your dream job?
A: I am living my dream right now. At Merlinos & Associates, I get to do regulatory work to help protect consumers and develop actuarial models for traditional insurance products. I also build predictive models for life, health and P&C insurance, and other industries like banking, credit rating agencies and other companies needing data-driven solutions. I get invited to speak and write papers on the subject of predictive analytics. I serve on the Predictive Analytics and Futurism section of the Society of Actuaries (SOA) and will be organizing the involvement of the section to present on predictive analytics at this year’s Valuation Actuary Symposium. The predictive analytics mini-track I organized last year was strongly attended, and the expectation is that this year’s mini-track, which will focus on incorporating behaviorism in predictive analytics modeling, will draw even more participation. Make sure you attend!