In the never-ending quest for pre-qualified leads, increasing numbers of insurance companies are turning to predictive analytics to net potential customers from that great sea of people known as the general population.
Popularized in the 2011 movie, “Moneyball,” predictive analytics enabled the salary-poor Oakland Athletics to nearly win the World Series by hiring low-cost players who intuitively seemed like long shots — but statistically were great prospects. Essentially, the A’s — like many insurance companies these days — were able to uncover hidden insights about offbeat prospects by finding surprising patterns in their stats with the help of computer algorithms. Like many organizations, the A’s had been awash in all sorts of player statistics for decades. But with predictive analytics, they were able to get fresh perspectives on those numbers, and then go on to beat teams with much deeper pockets.
“Organizations are already struggling with too much data,” says Simon Gao, vice president, TruStage consumer analytics, a brand owned by CUNA Mutual Group, a mutual insurance holding company that markets life and other insurance to scores of credit unions. “Predictive analytics is the most powerful solution to the ‘too much data’ problem.’”
In the life and health insurance industry, predictive analytics service providers generally start by studying the characteristics of people who have already purchased a product from an insurer, and then develop a profile — or model — of the kind of person who buys that specific insurance product. Not surprisingly, much of the profile that develops often seems like common sense. Many of the same people who are in the same age bracket, same income bracket and who live in the same kinds of neighborhoods often buy the same kinds of insurance products.
But the wizards of predictive analytics also say that some surprising patterns often pop up once all the data is crunched. “We recently did a customer segmentation study for a health insurance company and found some interesting results,” says Grant Stanley, CEO, Contemporary Analysis, a predictive analytics service provider.
To wit: Stanley found that when men are the primary health care providers in their homes, their families tend to use more preventative health care, and make fewer visits to the emergency room. Conversely, when women are the primary health care providers, Stanley’s predictive analytics found that the family relies less on preventative health care and more on the emergency room. Essentially, the finding flies in the face of conventional wisdom, which generally concludes women tend to be more focused on health care than men, and that men need to be dragged to doctors’ offices — especially when it’s preventive health care. Moreover, Stanley also found that once kids grow up and leave the house, there is no statistical difference in the reliance on health care between men and women. With an empty nest, both men and women access health care with the same frequency.
This insight saved Contemporary Analytics’ client of lot of money, Stanley says, since the health insurance company only needed to design a single marketing campaign to reach out to both empty nesters. The insurer, he says, “no longer needed to create two unique marketing campaigns.”
But perhaps most startling was that young bachelors turned out to be more loyal to their health care provider than were young bachelorettes. “Who would have thought that a 22-year-old man is more loyal than a 22-year-old girl?” Stanley says. Of course, the profiles — or models — a predictive analytics service provider comes up with once a company’s database has been worked over will vary, given the fact that no two companies or no two insurance products are exactly alike. But once an analytics provider has that profile, he or she can use all of its features to search for people in a general population that match all — or at least most — of the profile’s features.
As insurers have been doing for decades, predictive analytics providers turn to consumer characteristic/behavior databases that are maintained by credit card companies, Internet advertising brokers, the U.S. Census Bureau and similar entities to make those matches.
“Most marketers can arbitrarily group customers,” but that’s not predictive analytics, Stanley says. Predictive analytics goes well beyond arbitrary segmentation by ferreting out the statistically significant reasons — the true reasons — why customers behave differently.
Adds says Christopher Coloian, Predilytics CEO, another predictive analytics service provider: “In a city the size of New York City, for example, it’s possible we could identify tens of thousands of potential customers for a health care company.” Of course, predictive analytics is no panacea. Granted, the resulting list of potential customers for an insurance company generally features many potential customers who truly are interested in the insurer’s products. However, not every name on the list is a hit. Anyone who has rented movies with Netflix for any amount of time knows this. Netflix has been using predictive analytics software for a number of years now to recommend the kind of new movie titles you’d like, based on the titles you previously rated as pleasing. And yes, sometimes, Netflix hits the bulls-eye with such guesses. But other times, people are left wondering, ‘What was the computer thinking?”
Even so, TruStage’s Gao says he has had great luck using predictive analytics for his own purposes: to uncover the most likely of credit union customers for its health products. “We test various models to identify what message — or combination of messages — is the best for a particular segment of consumers,” Gao says. Essentially, the tests identify the “winning approaches with the biggest impacts.”