Call it confidence or call it an executive mandate, but the arguments have ended and the verdict has come in on the role of analytics and big data. Eighty-three percent of all P&C insurers plan to increase spending on data and analytics over the next three years, according to data from insurance-focused strategic advisory firm Strategy Meets Action. At a time when IT budgets among insurers barely budge, 43 percent of the same group plan to increase data/analytic spending by at least 6 percent per year and almost one in five plan annual increases of 10 percent or more.
“There’s finally an acknowledgement, especially in the really big companies, that big data analytics really will create competitive differentiation from pricing to selling to servicing,” says Deb Smallwood, the founder of Strategy Meets Action (SMA), a research and consulting firm.
Because analytics is more relevant throughout the enterprise, senior business and IT leaders are confronting more and differing opinions about the potential payoffs, as well as how to build an analytics culture. Plus, there are the typical change management challenges. On one side are entrenched practices and roles with often stubborn executors. On the other side, big data analytics beckons with the need for attendant skills at the very moment IT is hard pressed to staff current obligations for data repositories and reporting.
“It used to be that to be mature in data management you had a strong data architect, a leader in the IT organization, and that was that. With big data, new data sources and predictive modeling, insurers are shifting to the data scientist,” says Martina Conlon, principal at Novarica, a research and advisory firm for the insurance and banking industries. Though loosely defined, it’s the data scientist who straddles the boundaries of traditional and new data sources. “The data scientist is the person with a quant and IT background and domain expertise,” Conlon says.
The payoff for insurers is better statistical forecasting via predictive models, explains Heather Wilson, the chief data officer at AIG. “We’re investing in this knowing we already are a risk data organization, so it’s not new,” Wilson says. “We already do A through G, so let’s just start at H and move on with IT infrastructure and analytical tools. We’ve got to leapfrog, we haven’t done things in insurance that other industries have done.”
High-level mandates for change can sound like lip service for the task at hand, but change is taking shape. Through better tools and strategies and new roles, especially for data scientists, leadership is slowly but inexorably enforcing its will. This mindset was plainly on display at the April Insurance Data Management Association conference in Philadelphia, where senior data managers from AIG, Travelers, Chubb and others, aired their agendas for 2014 and beyond.
RETOOLING FOR BIG DATA
Jorge Rosas, VP of operational analytics, field operations services division at Chubb, says the new skills in operational management are engineering based. “Industrial engineers bring along operational research and management science. That kind of background has proven very useful in being able to communicate within the data science with the modelers.” Along with a shift away from the MBA recruiting, Rojas says Chubb is placing data scientists globally to model and explain specific markets previously undifferentiated except by financials.
The largest insurers are said to have as many as 60 data scientists on staff. “Midsize and smaller insurers that want to develop predictive models or leverage big data and analytics are more likely partnering with consultants to provide the expertise,” Conlon says.
Leaders from several firms stressed that they are not trying to reinvent their business, but rather would put them on a better learning curve. Erik Roen, VP of the claim business, intelligence & analytics at Travelers, says big data comes with many implications for the business. “Before putting a model into production across the entire claims organization, we want to get comfortable and learn things first,” he says. “It requires a process and competency on the back end making sure IT is at the table to manage the deployment and monitor it.”
Another process enabler is the rollout of powerful visualization tools, but Roen says there’s a fine line between seeing something interesting and making a false inference from the data. Travelers has an analytic group with a tacit knowledge of the business that helps claims workers explore more data, but makes sure they bring it back for regression or other analysis, he explains. “We’re not asking them to build models, but we are asking them to make sure they are asking some key questions so they don’t make a bad decision based on a partial picture.”
Nonetheless Travelers now is gathering insight from 500 million unstructured claims notes now aggregated in a searchable database and being tested in predictive models. Likewise, as it ponders a direct-to-consumer business, the company is performing a better clickstream analysis of consumers who act on Travelers’ 1.5 billion monthly web ad placements. “We can now process that in a very efficient and sequenced way for our marketing folks to get almost real-time looks at who is clicking through, where they’re stopping, and fine-tuning how we are placing ads based on feedback,” Roen says. “We couldn’t do that even a few weeks ago.”
According to a Novarica study of 55 insurance CIOs, the focus areas where big data is delivering either “significant business value” or “some business value” to at least 30 percent of those surveyed include actuarial/product marketing, underwriting and marketing. Almost 60 percent of the sample will be using big data for actuarial/product marketing and marketing in the next 12 months.
“With big data, insurers are engaging in a limited fashion, most commonly with consumer, business and geospatial data for benefits in underwriting, actuarial and product development,” Conlon says. “Insurers are utilizing what I’d call small-data analytics much more broadly, but many are planning big data initiatives over the next 12 months.”
What’s new is that the business leaders are trying to use data and analytics better by looking forward as well as backward, explains David Helmuth, a specialist leader in insurance and information management at Deloitte Consulting. “In the past it’s been our actuaries and our underwriters who have been most successful with data, so let’s bring them to the table and engage them in order to get traction across the enterprise and be more data driven.”
Novarica’s study also finds that leveraging analytics has been hamstrung by fragmented data environments, lack of attention due to other priorities, lack of investment or sponsorship and inexperienced staff.
With control as well as expediency on their agenda, senior data managers usually like to manage fewer repositories of data rather than more, which might otherwise sound counter intuitive. “We had so much wild, wild west going on,” AIG’s Wilson says. “We put a blueprint out there, got the buy in from respective CIOs and CEOs and said, okay, do you see the players on the field? We’re only going to have 18 authoritative data sources: that means those other thousand things you think are sources of data are going away.”
CHANGE MANAGEMENT
If technology and infrastructure are significant but known challenges, affecting process execution in areas of underwriting, claims, actuarial and others is likely going to be harder. Change management is much more serious business than business leaders usually take it for, according to Kevin Toth, VP of Harleysville, a Nationwide Insurance company. “I believe change management is one of the single most important and single most undervalued tools available to management today,” he says. “Change management is about preparing people to succeed in a new environment. It takes case building, relentless communication and constant attention.”
Internal and external change are separate but important components of a data-driven culture, Toth says, when an organization is planning a monumental leap from 20th century underwriting to underwriting that’s much more heavily informed by analytics. He describes the three common archetypes that stand in the way of internal change management. The first is the rebel, outspoken and seething at change. The rebels are entrenched and dangerous because they win converts and build armies. The second archetype is the ostrich, which tends to be abhorrent of predictive model rollouts. This is the person who wants to keep their head down and hopes the world will not have changed when they come back. Last and hardest to deal with is the maliciously obedient type who takes an initiative that’s supposed to be evolutionary and implements it as if it is revolutionary. These are the types who over-execute with a degree of passive aggressiveness, Toth says.
Predictive models seldom are rolled out as ruthless decision-making tools, but rather are intended as tools to become comfortable with. “The maliciously obedient person follows the model every time,” Toth says. “They can be the nicest people, but you end up having to sit down with them and explain that the plan to use the analytics isn’t hard policy; that you’re more interested in what parts of the model they are comfortable with and what parts they are not.”
Generally, opponents of change are best dealt with by anticipating them, communicating early and often so resistance can’t build momentum, and engaging with their concerns, Toth says. “You’re not going to win over everybody, but the goal is to stop them from multiplying. Give people another view so those on the fence can make up their own minds.” Most organizations don’t put managers in positions to affect change, and then not give them the necessary tools. “They assume that all this people stuff is going to take care of itself, but it doesn’t,” Toth says.
FUTURE INCLUDES AUTOMATION
In the age of mobile apps, mashups and tablet computing, it’s likely analytics increasingly will be embedded within workflows and will automate some of the changes management is trying to bring about.
The success of a model can be highlighted by embedding it in workflows and showing underwriters the results produced by the models, compared to previous methods. “I’ve seen some insurers make it optional for the underwriter to use the model, but if you don’t you must make a note in the file as to why you didn’t,” Smallwood says. Others may require home office approval if the model is not used.
“It comes together in the process,” Helmuth says. “It’s not necessarily just about assembling the underwriting for a risk assessment; it’s also the workflow around producing the underwriting report. There will be elements that are defaulted from a tool or dashboard to get the right information onto that platform.”
Moving analytics upstream to where the data lives is part of that future. “You’re really asking about the idea of embedding statistical models into your Massively Parallel Processing and Hadoop platforms and filtering there,” says Suresh Selvarangan, another Deloitte insurance specialist. “You move the model to the analytics. That idea is taking off, and big vendors are banking on it as the next big thing.”
Today’s focus though remains on recruiting, motivating and keeping data scientists around to build the connections to the future, yet another challenge for industry leaders. “I remind leadership that insurance companies offer more limited options for data scientists, especially midsized or smaller firms,” Conlon says. “Certainly you could have very fulfilling careers, but there are 15 or 20 different options in insurance, and data science is a very open-ended career right now.”
The industry has a job ahead in order to retain these experts and keep them challenged with their career, Conlon says. But above all else, those who eventually profit most from analytics and big data will be those who instill a culture where leaders trust analytics and act on their insights. “It’s a matter of time, you need to create that culture today if you don’t already have it,” Conlon says.
Jim Ericson is an independent consultant and award-winning journalist who writes extensively about data and information management topics.
