Login Register Subscribe
Current Issue

Big insights out of big data

Reprints

Big data in the insurance industry, which is used to reveal patterns and trends, looks poised to grow even more.

“The term "big data' has been a buzzword in the industry for some time now,” said Allen Narkiewicz, a partner at KPMG Management Consulting, Chicago. “There's more data in the insurance industry and outside of it — social media and other places — than there's ever been before.”

Big data and what it can mean to insurance companies “is utilizing data to drive fact-based data decisions. A big data (analysis) done right has to start with, "What problem are you trying to solve?' The factors that go into it are completely dependent upon the subject matter,” he said.

A big data analysis not done well “is just getting lots of data and messing around with it,” Mr. Narkiewicz said.

“Historically, insurance companies focused on, and were pretty good at bringing together, the data within their four walls,” said Michael Adler, a New York-based principal in KPMG's insurance advisory practice. “Now it's all about bringing in all this other, richer information — whether it be social media, whether it be around wearable devices, the internet of things, unstructured content — to help make better decisions, not just reactively but proactively and predictively.”

A 2015 study by the consultancy Infosys Ltd. said insurance companies can use big data analysis to manage risks profitably, identify fraud effectively, prevent identity theft, design products for customers, and acquire new and retain existing customers.

Adam Cottini, managing director of the cyber liability practice for Arthur J. Gallagher & Co. in New York, said big data is used to look at such concepts as buying trends and catastrophic risk.

“In the world of cyber that we live in, you're looking at data analytics to determine if there's exposure to aggregation, all of which can be used in determining the extent of risk and how your risk is spread across a population,” Mr. Cottini said.

Anand Rao, principal in PricewaterhouseCoopers L.L.P.'s advisory practice in Boston, described big data as the three Vs: volume, variety and velocity.

“The reason they call it "big' is because, every second, we're producing a whole lot of data,” he said. “You need to make sense out of that and then look at how you can use that data to maybe minimize losses.”

For Libby Christman, vice president of risk management at Ahold USA Inc., it's all about getting answers.

Like most retailers, the majority of the Carlisle, Pennsylvania-based supermarket chain's litigation is from slips and falls. “We're using data to help make decisions on safety programs by helping us pinpoint root causes of injury injuries and accidents based on severity,” she said. “It could be a problem with how are we reducing injuries, how are we looking at our cost lines to reduce them, but we start with a problem and ... a question about what data may be able to pinpoint a solution, or at least a starting point.”

Ms. Christman said the amount of available data “can be very overwhelming” but informative.

“I think it's helped us make much better decisions about where we use our investments in terms of our safety resources and our personnel,” she said. “I think we're able to tell a better story to our operations leadership by using data that's a little bit more pointed about what their problems are and what we need to do to fix them.”

Ahold's next big data project will review litigation data and attempt to highlight patterns.

“As we continue to get more refined in the information that we're getting, we'll continue to take on other projects,” she said.

Ben Fidlow, New York-based global head of core analytics at Willis Towers Watson P.L.C., said insurance brokers use data mainly “to provide decision support for our clients and differentiate our services.”

“We're all about empowering our clients,” he said. “If we can perform some type of analysis which allows our clients to make better decisions, we will go about doing that.”

Mr. Fidlow said more work needs to be done in predictive analytics. “Making that linkage from the raw data to the predictive aspects is something we need to get better at,” Mr. Fidlow said.