For P&C insurers, predictive modeling and analytics offer the promise of greater profitability and operational efficiency in an environment that is increasingly competitive, commoditized and characterized by persistently low interest rates. And so, the need for predictive modeling talent is growing, and the competition for talent - from inside and outside the industry - is intensifying.
Insurers derive a host of benefits from predictive models and analytics, and usage is pervasive and growing aggressively. In fact, 82 percent of insurers now use predictive modeling in one or more line of business, more than 40 percent now use analytics for pricing and rating, and 20 percent for underwriting, according to a joint study from Earnix, an analytics software provider, and ISO, a supplier of statistical, actuarial and underwriting claims information. Ranked by survey participants, benefits include increased profitability, cited by 85 percent, followed by risk reduction (55 percent); revenue growth (52 percent); and better operational efficiency (39 percent).
But the challenges insurers face in their efforts to recruit and keep the people responsible for building the predictive models to drive those improvements, are formidable. Almost half of the survey participants (47 percent) said they don't have enough skilled modelers. Other hurdles include a lack of know-how/technical expertise (42 percent). Culture (16 percent) and change management (15 percent) also make it difficult for insurers to hire and retain predictive modeling and analytics talent.
Allstate Corp. currently has between 50 and 60 statisticians on staff and is looking to hire another 20, for example. "For 20 positions, that's probably three good years of hiring," says Eric Huls, Allstate's SVP of quantitative research and analytics. "And that assumes no attrition of current folks, and no subsequent increase in demand, so it's a daunting challenge in terms of getting to full staff. By the time we get there, we will be looking to hire even more than that."
Until recently predictive modeling positions were not sought after on such a large scale. But with the maturation of the technology and the huge amount of data now available to companies, there is a growing demand for data scientists and modelers, and a short supply of talent, according to Jacobson Group, which offers insurance staffing, temporary staffing and executive search services to the insurance industry.
"Demand has been rising steeply over the past 24 months," says Brad Whatley, SVP at the Jacobson Group.
According to the last "Insurance Labor Outlook Study," a joint report from The Jacobson Group and Ward Group, which offers benchmarking and best practices studies to insurance companies, analytics surpassed even actuarial in terms of positions most in demand industry-wide.
During the past six months, Jacobson Group said, there has been a 63 percent increase in demand for big data jobs, including predictive modelers. And Gartner Group predicts there will be 1.9 million new big data positions created in the United States, and 4.4 million globally, by 2015.
As insurers expand their use of predictive models to personalize products, services and customer interactions, they increasingly are competing with other "sexier" industries, such as online retailing and other financial services industries for analytics talent. To satisfy their need for analytics talent, insurers, such as Allstate, CNA and Insureon, are finding they need to reconsider the hierarchical corporate structures that are common to the industry, recruit modelers based on skills rather than experience and then work hard to keep them engaged.
Allstate: Casting a Wide Net
At Allstate, the product research department used to concentrate on understanding the likelihood a customer was going to have a claim and how much that claim would cost, explains Huls. "But the breadth of the types of problems we are tying to solve continues to broaden," he says. "We also are getting into marketing and sales processes," he says, and applications now include discerning the likelihood of a customer to buy a product if one is offered to them, the likelihood they will renew their products, and their likelihood to buy additional products, for example. "We have built models to understand which claims are likely to be fraudulent. We build models around operational things, like which homes we should inspect before we insure them, because they might have condition issues," Huls says.
Those increasing opportunities to improve performance and drive revenue also are driving the need for more predictive modeling talent and helping create the unbalance between supply and demand for talent, Huls says. To help alleviate the need for talent, Allstate casts a wide net, Huls says, and looks for statistical and analytical skills, rather than insurance or other professional experience, which is different from recruiting for other positions, he adds.
To hire 20 modelers likely will require several hundred interviews and phone screens, plus home-office interviews, Huls says, as identifying technical skills on a resume is only the beginning. "We need folks who are able to apply those skills in a business context, working with partners both to understand the business problem that becomes the math problem and then translating the mathematical solution into an implementable business solution. Finding all those abilities in one individual is difficult," Huls says. Even then, a certain percentage of offers to desirable candidates will be declined, Huls explains, because talented people often can choose between offers.
"It's a numbers game in terms of talking to enough people to get enough qualified folks to come here," Huls says. "The top talent, or what could have been the top talent over the past two decades, has gone into business and finance instead of the hard quantitative stuff," which didn't pay nearly as well at the time, he says. "As these jobs get more press and salaries rise, we could see a shift back in what the universities are producing."
That doesn't mean that modelers can be complacent, however. "Many best-in-class techniques that we are using these days didn't exist or were not practical three or four years ago," Huls says, so keeping current with technology and continuing to hone their data and analytical skills is crucial for modelers.
Candidates must be able to do the math, Huls says, but their value comes from their ability to translate business problems into math problems that can be solved, and then partnering with business people to implement solutions. Intellectual curiosity is crucial, Huls says, as it advances the company's understanding and ability to act on data-driven decisions. "If we find people like that, we find they can pick up the business side of things. It's such a limited pool of candidates that when you start putting a 'has insurance experience' filter on it, it's really tough to find those people."
Allstate recruits actively from the University of Chicago, Northwestern and DePaul, and new hires tend to be recent graduates with advanced degrees in statistics, pure math or physics. Many assumed they would pursue careers in academia, but then decided to do something more applied. "There are people who spent years in Ph.D. programs, and who do truly ground-breaking research. But that research is of such narrow interest," Huls says. "They might produce a thesis or dissertation that is read by literally dozens of people after years of work. People are seeing that the work Allstate offers is no less interesting, but is much more impactful and tangible."
The next challenge for insurers is keeping them, as they receive multiple calls per week from recruiters.
"Whenever new and different ways of doing things arise, there is always the risk of creating friction." He says insurers have to make certain the work is interesting and impactful, and that it gets used. "The transition to analytic decision making really is an opportunity at Allstate, and many places, to reinvent how the business is run," Huls says. "That's an incredibly exciting thing for people to be a part of. Getting things in place so they can contribute is critically important, because if they don't have that line of sight, and they can't see how the work is making a difference, there are a lot of other places waiting to show them that they can do it there."
CNA: Training and Cross Training
At CNA, a commercial P&C insurer, predictive modeling is used for pricing differentiation, segmentation of loss costs, understanding risk characteristics, and granular pricing of risk, explains Renee Davis, SVP and a senior actuarial officer.
"That's been a key focus for us, product by product. But we also have efforts to support claim/resource alignment; getting the right types of claims to the right type of claim handler quickly," Davis says. "The quicker we can get a case into the hands of someone who understands the injury or the type of claim better, the faster things can be resolved at fair and reasonable settlements," which increases profitability and customer satisfaction, Davis says.
CNA currently has fewer than 20 dedicated modelers, Davis says, but the company has actively been building its analytics staff for about five years. CNA has responded to the talent challenge from several angles, she says, including active on-campus recruiting, working to understand and shape the content of academic programs and building relationships with the students.
But CNA also is ramping up cross training and continuing education efforts, Davis says. Non-statisticians actively participate in building project plans and, as the ultimate users of the models, have an important say about what goes into them.
"Our actuarial training isn't robust enough to jump into these high-end statistics," Davis says. "But we are starting to cross train and bring actuaries into the analytics group and get them involved. I see them merging, not diverging, but that's only going to be possible in a big company that is willing to dedicate time to do that."
When recruiting modelers, Davis says CNA looks for math and data skills but, more important, the company looks for modelers who want to understand the commercial P&C business.
"It's bigger than the math. The model has to be explainable," Davis says. "If a non-statistician is going to be using these models - an underwriter, let's say - if they don't understand it, they are not going to use it. What good is it if it doesn't get used? At the risk of being a little blasphemous, a 90-percent model that gets used 100 percent of the time is better than a 100-percent model that gets used 50 percent of the time."
The majority of modelers at CNA have master's or Ph.D.s in statistics, Davis says. There are few of those people available and, until recently, the expected career path for them was narrow and typically was limited to pharmaceuticals or academia, she says. Now their options are much wider. "It's pretty clear the path an actuary is going to take. They're going to go work for an insurance company or a consulting company. That's the focus. With these folks, every industry is looking for people with statistical backgrounds. Think about retail, the Amazons of the world and other high-tech companies. I've seen an article recently about HR companies using statistics to find talent. That's why the competitive landscape is pretty tough. There's a much broader set of opportunities for them," Davis says.
To increase CNA's visibility and build interest in the company, CNA offers internships to junior candidates, in addition to those who soon will graduate, to expose them to the work and the business of insurance. It also sponsors university classes and information sessions about how their skills could be deployed in the insurance industry.
"We are offering something beyond just modeling at a computer," Davis says. "We think that makes for a better and more interesting job, frankly. And we think that has longevity when people feel that we've invested in them. It helps us retain people."
Insureon: True Believers
Insureon, a Web-based specialty P&C insurer, has spent 12 years building a web-based policy administration system using .NET and C#, explains Ted Devine, CEO. The system is built on five million lines of code and uses predictive models and analytics constantly and in real time to drive decisions that incrementally speed the process of applying for, quoting and binding policies.
"We have a huge rules engine in the back that helps with classification. And it learns. It has to, to be able to deliver 550 verticals across four products across 50 states," Devine says. "It's a complicated thing."
The company has five predictive modelers on staff, all graduates of the University of Chicago or Northwestern University. And they are young, 23 or 24 years old, Devine says, each with degrees in statistics or finance, and most can program in SQL. Being in Chicago, which hosts many universities and attracts a lot of young people from neighboring states, offers his company a distinct advantage when it comes to recruiting predictive modeling talent, Devine says.
"We're a tech company; we're growing; and we're doing something that's pretty cool that's changing an industry. And we have a very flat structure. These guys are in conference rooms every single day talking about analytics and how we're going to change the business. They're not five levels down in a huge organization."
Devine has a history of leading in very flat organizations, and that flat structure, he says, is required for aggressively innovative companies. "I grew up at McKinsey, which is about as non-hierarchical a firm as there is. Through my whole career, I've had people in their early 20s who are smarter than I am, pushing and driving and making decisions. And that is what we want them to do."
If they were not involved in decision-making and problem solving, Devine says they would simply leave. "I'm smart enough not to do that. To make partner at McKinsey, you've got to empower the younger people, and then they follow you through the firm. That's a lesson I learned 15 years ago, so we've always had that mindset."
Devine says to keep that talent, and keep them engaged, he takes personal interest in their development, and frequently gives them new challenges. For example, Insureon's the new head of online marketing was until recently head of product, Devine says. "He doesn't know anything about online marketing - not a thing. A year from now, he'll know ten times as much as I do, because he's really smart and he's going to dig in and he's going to learn it. You can keep them as long as you can keep them interested. And that means moving them around in different jobs."
In addition to offering new challenges, Devine says culture is very important. "You can't recruit for it if you don't believe in it. I'm a stats guy from Chicago. I mean, I love this [stuff]. This [stuff] fascinates me. For me to sit with a 23-year-old and talk to them about what we need to do, and why it's important, and then work beside them to develop the products? It's cool for them and it's cool for me. The one thing that the C-level executives in insurance companies have to understand is, they need to be engaged in the leadership of these analytical thinkers. And if they are, then they'll get unbelievable amounts of value out of them. If they just hire them and put them in a corner and come see them once a year, that guy is not going to be there."
"You can do this in a big company," Devine says, adding that he brought this sort of creative culture and flat structure to Aon Risk Services, where he was president. "It's the power of analytics that counters traditional thinking. Yes, it's intimidating because a lot of people grow up and they use experience as their source of knowledge, which is very important. But at the end of the day, the numbers are the frickin' numbers, man. You've got to be willing to look at it and let it change your mind. So, yes, people get threatened because they rely only on experience to win the argument and these kids don't really care about that. They care about the figures and the model and what it says and then what the implication is. You get them really engaged, then they can change the world."
Chubb Introduces Disruption
At Chubb Insurance, there is a strong impetus to implement predictive modeling and analytics throughout the organization, explains Upendra Belhe, Ph.D., chief enterprise business analytics scientist. But hiring talent for predictive modeling and analytics is different from hiring other types of talent, he says.
Talent is scarce, the need is great and the competition is intense, but having the mathematical skills and a degree isn't quite enough to land a job on Belhe's team. At Chubb, the need to acquire talent is tempered by Belhe's insistence on a strict set of soft skills, including common sense, empathy, maturity and patience.
"People are quickly beginning to understand that taking the data and predicting into the future is only a small part of the game," Belhe says. "The bigger part of the game is story-telling, and I think more and more stories have been written and are being written with this data."
In addition to the ability to conduct statistical analysis or build predictive models, Belhe says candidates also need to demonstrate the right level of empathy for the business users who will use the models. Because predictive modeling and analytics are such disruptive technologies, Belhe says, they will put many business assumptions at risk. That, combined with the insurance industry's culture, which tends to be more conservative and hierarchical, makes maturity, which includes humility and the ability to listen, an essential ingredient for determining the candidate's compatibility with the team and the company, he says. - Chris McMahon
For more about Chubb's approach to recruiting modeling and analytics talent, go to http://bit.ly/LxlZLK
This Year's Model
Predictive modelers and data scientists manage data assets by strategically designing and implementing data warehouses, data marts and data stores intended to deliver high levels of data availability and integrity, Jacobson Group says. That helps create an information edge to increase delivery, support, and communication, driving profitability and growth.
Their educational backgrounds typically include majors in computer science and information systems, analytics, higher math, statistics, actuarial, risk management, physics, insurance, economics and engineering.
Many roles require a master's degree or Ph.D.
Salaries for predictive modelers vary by geography, responsibility and experience, and are not yet tracked by the U.S. Labor Department.However, the median pay for statisticians was $75,560 per year in 2012 and for actuaries was $93,680 per year.
According to an April 2013 story in The New York Times, the average salary for the 84 people who graduated from the Institute for Advanced Analytics at North Carolina State University in 2012 was between $89,100 and more than $100,000 per year for those with prior work experience; and all of those graduates had job offers.
For the insurance industry, salaries generally are lower, Jacobson Group says, but they are rising.
"For the insurance industry, this number is really dependent on the extent and breadth of the candidate's experience, the size and scope of the company and even the development stage of the employer's analytics program," Jacobson Group says.
"As the demand for this type of talent continues to intensify, industry organizations will need to build an innovative and attractive compensation package to compete for A-level talent," Jacobson Group says.
TOP 5 RECRUITING TIPS:
1. Broaden the search to other industries.
2. Demonstrate stability of the analytics program through executive sponsorship, ideally by leaders in the C-suite.
3. Share an exciting vision and an analytics story to create a recruiting edge. What has the company done, what are they doing now and what is the desired future state? Promote your company's operational excellence, innovation and the role of analytics in achieving success.
4. Highlight a dynamic, innovative and state-of-the-art analytics career path that will prove opportunities exist to use candidate's highly specialized skills to the fullest extent.
5. Emphasize the soft skills that separate a great candidate from business leaders, including the ability to build relationships and translate technical information into business objectives and bottom line impact. - Chris McMahon
