No doubt to the vexation of carriers and vendors alike, business intelligence (BI) and predictive analytics are often lumped together. Given the rapid growth of the technologies in recent years, the confusion is understandable, especially when one considers that both technologies use the wealth of historical and third-party data insurers have on hand to build models to either augment human decision-making or eliminate it altogether through automation.

Despite this cosmetic resemblance, the differences are profound. Historically, BI has become synonymous with operational metrics and monitoring, querying and reporting functions. Modern BI solutions generate scorecards, a collection of metrics and reports on a unified interface used to measure against objectives and dashboards, which use metrics to give users the pulse of an organization. While BI is reactive, and looks backward to gauge performance, predictive analytics seeks to use data in real time for sub-second decisions to affect future performance. While BI tools enable slicing and dicing and give insurers a high-level view of what’s going on, predictive analytics promises insurers actionable knowledge and a granular view of their operations. Thus, if BI is a look in the rearview mirror, predictive analytics is the view out the windshield.

“It’s not just analyzing, finding something interesting and wondering what you can do with it,” says Colin Shearer, SVP for market strategy for Chicago-based SPSS Inc., supplier of predictive analytic solutions, “it’s the ability to go seamlessly from analyzing historical data, finding hard facts, then deploying them to improve your processes. We can embed our models live in an operational process and in the systems that support them.”


Yet, whether considered separately or lumped together, business intelligence and predictive analytics are no longer seen as optional and are now established as a cost of doing business, especially in areas such as personal auto.

“For operational decisions, such as underwriting and claims, use of business intelligence and predictive analytics is going up rapidly and ubiquitously,” says Mark Gorman, principal of St. Paul, Minn.-based Mark B. Gorman & Associates LLC. “It’s no longer a simple ROI discussion, it’s no longer something you can put off until later—if you are going to be in the market, you are going to have to do it.”

Gorman says there are several factors for the expanded use of the technologies ranging from industry demographics to broad acceptance from upper management. “It’s well established now among actuaries and senior managers that this methodology and process is the wave of future,” he says. “This is a trend, not a fad.”

A second, perhaps more important trend Gorman sees is the technologies being used more for performance and strategic purposes. “We’re seeing more applications for strategic decisions about things like entering new territories or distribution channels,” he says.

Gorman isn’t alone in seeing the value of increasing BI deployment across the enterprise. Considering the ascendancy of service-oriented architectures and the advent of Web-service platforms that allow carriers to pull data from a variety of systems across the enterprise, it is not surprising that use of BI has spread apace.

“The value of BI is that it offers you consolidated analysis from disparate data sources,” says Craig Bedell, director of Global Insurance Services for Ottawa-based Cognos, a business intelligence company recently acquired by IBM. Bedell says that when it comes to data analysis it is best look at the pond, not at the stream feeding it. Thus, by aggregating data from across enterprise for analysis, the whole is indeed greater than sum of its parts, Bedell says. “Having BI as an enterprise capability, and using the insight you get from one department in another department, is key. If you can understand claims activity relative to agent or type of business, doesn’t it make sense to feed that type of info back to distribution management?”

One area were predictive analytics was quick to gain traction is in claims. Carriers can use analytics to assess prior claim history and then drive decisions within the claims process. Bill Dibble, SVP of claims at Birmingham, Ala.-based Infinity Property and Casualty Corp. says the company began to implement a predictive analytics solution from SPSS in October of last year.

Dibble says Infinity is using the solution to better categorize claims according to risk at first notice of loss, and route them accordingly. “Previously, loses were coming into a loss repair unit, then reassigned,” he says. “People were getting several adjusters on their claim. We pushed predictive analytics forward to model what type of claim should go out to our adjusting staff.”

As a result, he says, many smaller claims could be handled in house, over the phone by a loss-report taking unit. He says that currently about 4% of Infinity’s claims are being handled by the loss report taking unit, but, by the end of the year, he expects 25% of claims to be handled by that unit. Consequently, field adjusters, freed from working minor claims, can specialize more—and work on—only severe claims. “The customer has a better claim rep on his claim who can handle it in a much more effective manner, he says. It’s win/win for the company and the customer.”

Dibble notes getting these claims to the proper adjuster has already helped hold down costs in other areas as well.

“The biggest surprise is how quickly we were able to capture our return on investment for areas like subrogation,” he says, adding that new losses coming into the subrogation unit held steady at 15% or 16% for years prior to adopting predictive analytics, but immediately went up to 19% or 20% upon implementation. “The quicker I can get a claim into my subrogation unit, the quicker I can recover dollars. Usually there’s money going out, now there’s money coming in.”

Infinity is using a phased rollout to implement the solution. “A lot of this is dependent on your IT resources,” he says. “We’ve started out batch, but the real value will be when we can do it on a real-time basis and use analytics to help the adjuster ask the right questions.”

Another area where both BI and predictive analytics can pay rapid dividends is in fraud detection. Insurers have long used third-party data such as credit scores, to correlate with loss history. They also followed rules, such as, if an accident happens during the first 10 days of a policy, it deserves greater scrutiny. Carriers can now feed these variables into predictive models to estimate the possibility a claim is fraudulent before handing it over to special investigative unit. This approach gives hard evidence to investigators, and justification why the claims are suspicious, not just the intuition or gut instincts of claims adjusters.

Moreover, the visualization and link analysis tools common to BI solutions can highlight patterns that would escape the human eye.

“When you look at fraud in isolation, everything can appear to be above board, but when you look at it from a network perspective, you can see that some of these claims are related through, say, a common doctor or credit card account.” says Stuart Rose, global industry marketing manager, insurance for Cary, N.C.-based SAS Institute Inc., adding the tools allow investigators to further drill down into the numbers.


Just as predictive analytics can help weed out bad customers in the claims process, it also can assist carriers in rewarding good customers in the underwriting process. Such is the case at Pinnacol Assurance, according to David Sauther, team leader, decision support services, for the Denver-based workers’ compensation carrier.

Sauther says Pinnacol selected a predictive modeling tool from Denver-based Valen Technologies Inc. to improve an already well-functioning underwriting process. The Valen tool returns a score, which Pinnacol’s underwriters use to more accurately place their customers within a tier.

Sauther says not only the company is reaping the benefits of the improved risk segmentation, but so are its customers. Many of the small businesses that would have previously found themselves in the company’s default or standard tier, are now in the company’s superior or preferred plus tier, he says.

“One key difference is we now have tens of thousands of smaller businesses getting into our better tiers,” Sauther says, noting the differentiation between the tiers is now more distinct.

By adopting predictive analytics, the company wants to be able to underwrite small policies in much the same manner as large ones, which wasn’t feasible using manpower alone. Moreover, the technology helps underwriters discern minute distinctions between policies that, from a human perspective, would appear very similar.

“The value of predictive analytics is that we are able to take hundreds of data items and combine and weight them in a way no human mind could, then spit out a result that really tells more about the policyholder risk than you could just by looking at a few characteristics,” Sauther says.

What’s more, apart from some training for agency partners and underwriters, implementing the solution has not been very disruptive to the process, Sauther says.

“It’s just a different set of metrics and some weighting as far they are concerned,” he says. “They have no idea about the mathematics going on behind the scenes. The system generates a tier just like it used to, but now the tier is more accurate.”


Paradoxically, it is partly their ease with arcane mathematics that makes actuaries one of the latest converts to business intelligence. “BI solutions have been around a long time it’s a mature technology,” says Steve Goren, leader of New York-based Ernst & Young’s IAAS Actuarial Transformation practice. “If insurance companies are a little behind the curve, actuaries are even further behind the curve.”

Goren says that a reason for the relatively slow uptake is cultural. “Historically, actuarial departments are not well supported by IT, especially at smaller firms, and actuaries have been fairly successful doing reporting with spreadsheets,” he says. “In many cases, they’ll have hundred or thousands linked. Actuaries think ‘why send it to IT and wait two months when I can do it myself?’”

Yet, this go-it-alone attitude may be changing in actuarial departments, especially at large life insurers, where many are inculcating BI into valuation, modeling, planning and forecasting. Driving this is a growth in the amount of data, spurred in part by regulation.

Regulations such as the Sarbanes-Oxley Act (SOX), the forthcoming Solvency II regulations in the European Union and a tightening of the International Financial Reporting Standards by the International Accounting Standards Board, are driving this. As a consequence, enterprise risk models are being built for insurers to model this risk across the enterprise.

“What SOX and Solvency II have brought to the fore is that there is an interrelationship between risks and financial factors within an organization,” says Gorman, noting insurers should not look at risk within line of business, but rather across lines of business.

“What we’re seeing is that between Solvency II, IFRS and principal-based capital, they are starting to realize they’re going to need help from IT, even just to get the models to run in a reasonable amount of time,” Goren says. “With principal-based reserves and stochastic modeling, this volume of data is not going to be manageable in the spreadsheet world. With the tens of thousands scenarios actuaries have to run, it’s just too much data.”

Indeed, the volume of data actuaries will need to transfer from data warehouses to feed into their modeling or valuation engines will require them to make new friends in the IT department.

“As these data volumes grow, you are going to need to capture in an automated fashion the outputs from these engines in some sort of reporting repository and then put a BI solution on top of that,” Goren says. “A lot are moving to grid computing to get their valuation and modeling engines to complete calculations in a reasonable amount of time, because they recognize they can’t do it themselves.”


Yet, wrangling the data to feed business intelligence and predictive analysts is not merely a technical undertaking. Successful use of BI requires rigor on the part of insurers in their approach to the quality of the data in their information warehouse.

To ensure the quality and uniformity of data, metadata is vital for putting information into the proper context for analysis. “What we’re starting to see happen in the insurance market is a healthy recognition that data is still the fuel that runs the engines,” says Bedell. “You have to line things up, and the way to do it is with a metadata layer, which lets you line up apples with apples and oranges with oranges.”

Further complicating the picture is the rise in analysis of unstructured and semi-structured data. Carriers can text mine notes in claim files, for example, to search for patterns. “In my research, I’ve found that tremendous amount of data within an insurance company is unstructured,” Gorman says. “Turning that into some type of structured format for analysis, or using tools to analyze unstructured data, is clearly a trend as well.”

Infinity’s Dibble says the carrier is working on text mining everything from statements to adjusters’ notes. “Due to the vast workload of claims adjusters, they just don’t go back into their notes and correlate them with what’s happenings with the claim,” he says. “There’s valuable information within that text that can help us with our handling of claims, especially with the SIU area.”

Moreover, Dibble is bullish about the use of BI and predictive analytics spreading elsewhere in the enterprise—specifically in areas such as product management and underwriting. “I see this as having more than just a claim application,” he says. “We haven’t even scratched the surface of what it can do.”

Editor’s Note: As a continuing dialogue on this important topic, Insurance Networking News’ will present a Predictive Analytics and Business Intelligence Roundtable this fall—watch for details to follow.

For more on BI, search “Are Insurers Gobbling Up SOA-Enabled BI Tools?” at

(c) 2008 Insurance Networking News and SourceMedia, Inc. All Rights Reserved.

Register or login for access to this item and much more

All Digital Insurance content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access