One would think the insurance industry—the most data intensive of all businesses—would be leading all others in applying predictive analytics to better price and market its products and services. Think again.
The industry has long been criticized as a technology laggard, an insult it deserves. No longer, however, will this be the case. With customers’ expectations increasingly shaped by their experiences with digital powerhouses like Uber, Amazon and Facebook, the industry finally is investing some of its immense capital in predictive data analytics and modeling tools.
This is a very good move. It will result in a stronger industry, able to fend off the encroachment of technology companies eager to bite into the insurance pie. Data analytics also will drive greater efficiency, cost-effectiveness, better products, more loyal customers and smarter decisions regarding new markets and territories.
Five years ago, brokerages and carriers toiled in the backwoods of data analytics. Today, they are emerging from the darkness to combine their internal data repositories with external social, geopolitical and macroeconomic data to create actionable insight. It’s a significant trend. “The industry,” says Ari Libarikian, a director at McKinsey & Co., “is at an inflection point.”
Until now, most brokerages and carriers sought to improve their predictive capabilities via business intelligence software, which uses only internal operational and management data to create the basic metrics that help companies run better. This is not advanced analytics, as it excludes other sources of data that may play a role in determining, for example, potential claims for a product in a particular region.
“Internal data alone does not always predict or hypothesize the future, guiding better decisions on how to handle claims, underwrite policies or determine which customers to target and which markets or geographies to enter,” Libarikian says. “True data analytics takes into account other factors to arrive at a more assured decision. The industry has been slow to adopt this, but this is where it is headed.”
It’s About the Data
Predictive data analytics incorporates a variety of statistical techniques to analyze current and historical data to make predictions about the future. These techniques include modeling, machine learning and data mining.
All organizations have their own internal performance data. Analyzing these historical facts can indicate a range of possible outcomes. To make better predictions, organizations also need to access and analyze external data—the social, geopolitical, macroeconomic, climatological and other events in the world that can influence their decisions to develop a new product or enter a new market.
The insurance industry has been adept at assessing its internal structured data—the information residing in a company’s databases and traditional line-item spreadsheets, producing metrics on sales, customer demographics and other business indicators.
What’s been missing is the analysis of this internal data in combination with external data and (in a perfect world) unstructured data. The latter includes information that doesn’t fit gracefully inside a spreadsheet, stuff like corporate email, digital documents, photographs, Web pages, and video and audio files.
Once all this data is collected, advanced analytics involves applying sophisticated algorithms that carefully balance the relative importance of the data sets. Simply put, an algorithm is a set of problem-solving rules or instructions that lead to a predictable result. In other words, algorithms sift the wheat from the chaff. Out pops the predictions, which then assist more informed and confident decisions. Indeed, the more data a company can collect, distill and analyze, the better the opportunity to transform numbers into useful fodder for decision-making.
“Data increases in value at an exponential rate when you are able to add another related data set to the profile,” explains John Rizzuto, vice president of software, social media and information management for market research firm Gartner. “By having this information a millisecond ahead of your competitors, you are in a better position to encourage the consumer through pointed advertising messages to get them to do something.”
Insight is everything in business. It drives which product to make or improve, which market to enter or exit and which geographic territory to engage or avoid. Armed with predictive knowledge, operations can become more efficient and workforces more productive. Customers will be more satisfied and open to cross-selling and up-selling opportunities, and the organization’s controls will be stronger from a compliance standpoint.
Insight, then, represents the Holy Grail of business. Once captured by brokerages and insurers, it will transform the industry, differentiating organizations based on their competitive strengths and specialized skills.
Making Sense of Data
Many other industries already have figured this out and taken this direction—and for good reason. In today’s blisteringly paced global business environment, the organization with the fastest access to leading business opportunities and risks can act on this information more quickly and decisively.
Spoiling this picture for years has been the sheer overabundance of information, at least since the Internet dawned a generation ago. Businesses are overwhelmed in trying to make sense of political and economic data, weather patterns, customer demographics, competitor moves and even social and cultural trends.
What once was ink on a couple dozen paper documents is now digitized into millions of data sets, clogging what we used to call the Information Superhighway until it became a massive traffic jam. “There is just so much noise out there in terms of data,” says Doug Laney, research vice president at Gartner.
Fine-tuning the noise to the right frequency is the goal, but until recently it has been an elusive one.
“The challenge is to access and distill the huge volume of external data to make sense of its meaning in terms of your business,” says Rich Wagner, founder and CEO of Prevedere, a forecasting software company. “You can then correlate this data with your internal performance data into a set of leading business indicators.”
Laney concurs there is tremendous value in analyzing both internal data and external data—the vital stuff “that resides outside the four walls of your business,” he says. “The ability to curate the Internet’s storehouse of information and intersect it in unique ways with a company’s own proprietary information will separate the winners from the losers.”
Many industries understand this relationship. Banks and other financial institutions, such as stock brokerages, use advanced analytics to better understand their customers in relation to their particular products, distribution channels and rates. Based on this information, they will market products aligned with their customers’ progressive life stages—marriage, children, college, retirement.
Retailers also use data mining to forecast market trends, analyze purchasing behaviors, customize advertising messages, and charge a price that customers will be willing to pay. Both brick-and-mortar stores and online retailers are investing in technologies that accumulate information regarding a customer’s demographics, buying history, social media interactions and Internet search and browsing activities. These data are then blended with external social, market and economic data to generate personalized pitches to the consumer.
Another industry turning data into dollars is real estate. The online real estate firm Zillow, for example, is accessing previously unobtainable government information like census and permit data to make better predictions on the value of a house. President Obama issued an executive order to make the data freely available to the public in real time and in an easily understandable and accessible format. Previously it was unformatted, unstructured and out of date.
Zillow applied the new data to its other metrics on the market value of a home to improve its forecast accuracy. The result was remarkable—the company nearly cut in half its median error rate of a home’s final sale price, from 13.6% to 6.9%. In addition to the United States, governments in the United Kingdom and the European Union also have opened their data storehouses on health, education, worker safety and energy.
Manufacturers, on the other hand, are leveraging machine learning to improve asset utilization, manufacturing rates and product quality. By equipping multiple machines with sensors that measure oil temperature, water pressure, air pressure and equipment vibration, operators can achieve greater efficiency and productivity. Not only are fewer people needed to run and maintain the equipment, the machines also speed up or slow down based on customer demand data.
Municipalities also are leveraging machine learning to reduce costs. A case in point is California’s drought-stricken Santa Clarita City. Last year city officials retrofitted thousands of irrigation controllers with sensors to water more than 700 acres of parks and medians. The sensors measure ambient conditions such as moisture, heat and exposure to strong sunshine at different intervals of the day and season, in addition to the slope of the land, soil and plant type. Nineteen different data sets are analyzed using algorithms to determine optimum irrigation, helping the city save more than 20 billion gallons of water in the first year of the system’s operation. With restrictions recently imposed in the state, look for more of these high-tech solutions.
Predict the Future
These are just a few of the industries investing in data analytics to inform business decisions with quantitative assessments augmenting gut instinct—the experience of trained employees. Sitting on the fence while a competitor seizes upon advanced analytics to increase sales, improve marketing, cut costs or better price its products is no longer an option in any industry. As a recent Forrester Research report warns, “Firms that don’t modernize stagnate.”
Yet too many companies wait for others to make the first move. This is partly attributable to the technological inexperience of their CEOs and board directors. Libarikian cited a McKinsey survey that found that many boards rarely have technology-related discussions. For too long this was certainly the case at many insurance companies and brokerages, but the tide is turning. “We’re definitely seeing a shift to where most now fully recognize the potential of advanced analytics,” Libarikian says. “Our research indicates this is now a top-three or top-five CEO agenda item in the insurance business.”
Four years ago, predictive modeling was largely perceived by the industry as “an interesting concept that needed further study,” Libarikian adds. “There is much less of that kind of debate now, with high-performing companies separating themselves from the pack.”
Such organizations are launching advanced analytics programs and other advanced initiatives, Libarikian says, as well as hiring chief analytics officers and building teams around them. “There’s more model building, more experimentation and more insights coming out of these developments,” he says. “This is a far cry from traditional actuarial science.”
Much of the activity in the industry involves assessing claims data in relation to other data sets. For example:
- A property claim may be compared to a region’s building construction codes, weather and history of natural disasters, as well as the policyholder’s financial stability and business performance data.
- A workers compensation carrier might compare a client’s claims data to its workforce demographics, health and safety management program and financial condition—the latter represented by its stock price.
- A life insurer might analyze a prospective customer’s health history in relation to geo-demographic factors and treatment options for a particular disease.
This is just part of the power of predictive analytics. Here’s another: Say a carrier or brokerage is planning to enter a new geographic territory. A range of geopolitical, macroeconomic and regulatory data sets will inform the efficacy of that decision. The region’s currency might be in trouble or its interest rates might be headed sharply skyward. When combined with other data sets, this information could indicate that it might not be the best time to put a stake in the ground.
“Where predictive analytics gets really interesting is when you test variables you wouldn’t think are strongly correlated with risk, such as the tenure of the management team of a commercial client or the frequency of its mentions in the press,” Libarikian says. “One might not think these factors would have direct linkages to certain outcomes, but they may.”
The challenge is to carefully fine-tune the algorithms within the data analytics, weighing each of these variables according to their import. As Libarikian points out, “Often it is the relationship between the different variables that predicts the outcomes, not the variables themselves.”
The bottom line with data analytics is that it is simply a forecasting tool, albeit one that reduces the likelihood that a company will make a bad decision. For the most part, carriers and brokerages are just beginning to accumulate external data and turn unstructured bits of information like email and video into constructive information.
“Where the industry is going is to a very exciting place,” Libarikian says. “It will be difficult for an insurance company or brokerage to be successful five years from now without a strong advanced analytics capability.”