The (Business) Intelligence Community for Insurance 

Nationwide Insurance realized it had a wealth of data on interactions with customers, including records of compliments, questions, and complaints. What it didn’t have was a good way to put that information in the hands of people who could make a difference.

"We realized we weren’t doing a whole lot with that [customer contact] data, so we wanted to create a system that would offer real-time access to business intelligence and analytics to all our users and to provide usable and measurable data in an easy-to-use, Web-based environment," says George Armstrong, Nationwide’s director of corporate customer relations.

Nationwide captures customer feedback from any channel—phone calls, Internet transactions, postal mail, and social media—within a centralized customer contact database. Leveraging an investment in a BI platform already in use in the enterprise (which Nationwide declined to name), the carrier built a customer experience management system it deployed in
January 2010.

Users log on to the Web-based platform, which presents dashboards customized to their areas of responsibility. Problem areas are flagged, and users can drill down into those areas to identify the root cause and take action. Department-level dashboards roll up information and activity to managers who can ensure customer service issues are being handled by individual staff and create service-improvement action plans.

User-group testing of the system has Armstrong optimistic about the benefit to customer service. "We already can see positive impact based on the fact we have information out there," he says. "It gets complaints and inquiries in front of the right people."

Nationwide is one of many carriers that are targeting performance improvement by putting actionable intelligence in the hands of the people directly responsible for that performance in underwriting, claims, and other areas. Rather than requiring business staff to approach IT for every modification made to predefined management reports, these companies have worked not only to provide information to people when and where they need it but also to allow users control over BI tools themselves in order to perform iterative analyses and drill down into underlying data.

However, opening up broader access to BI creates some new challenges and risks for insurers as sensitive or proprietary data is freed from IT lock-down. Managing those risks requires identifying best practices around pushing control to business users.

"The question insurers need to answer is who can see what and who can do what," Carnahan says.

That includes establishing permission-based access around who has access to standard reporting vs. ad hoc reporting. "Insurers tend to limit ad hoc or ‘deep drill’ access to certain roles or certain types of information," Carnahan says. "Ideally, insurers should establish different dashboards for different users so it’s easy for those users to get to the right level of data needed to do their jobs."

For instance, claims representatives might have a dashboard that enables them to manage their own caseloads. While it shows how their performance compares with departmental standards, the dashboard doesn’t include caseload details for the entire team. Higher-level users don’t have individual case detail automatically on their dashboards but can drill into that detail as needed to address hotspots.

Nationwide’s customer experience management system provides that level of user-based customization. "The tool has controls around what data can get pushed to individual users and how far those users can drill into underlying data," Armstrong says.

In addition to technology-based controls, carriers need to establish business-based best practices around pushing additional control of BI. "Carriers need to be sure their operational processes are in place for data security," Carnahan says.

For example, there’s no benefit to restricting a level of system access for agents if a marketing rep turns around and provides that information to the agent. Carriers need to provide training not just on tools themselves but on broader issues of data protection.

The trend of pushing BI to the business user is clearly seen in predictive analytics, where underwriters, claims, and other line-level staff use analytics tools for purposes ranging from supporting individual case decisions to monitoring the portfolio of work for which they are accountable.

Some definitions of BI encompass only the more traditional functions of reporting and analysis, but Carnahan prefers a more comprehensive view that includes predictive analytics, as well. "I would define BI as anything that helps you use data to analyze information, make decisions, and analyze performance," she explains. "That includes data mining, benchmarking, text mining, all the way up to predictive analytics."

Society Insurance targeted more consistent decision-making and pricing of accounts when it put an analytics tool from Valen Technologies in the hands of underwriters but quickly found the impact of the BI platform extended much further.

"What surprised us was our underwriting staff has used the tool beyond the individual risk pricing and selection," says Rick Parks, Society’s COO. "They’ve applied it to book-of-business reviews to analyze how that book should be priced and why, compare how we have it priced, and pro-ject what the results could have been. We’re finding some unexpected opportunities."

Society contracted with Valen in 2008 and put the tool into production in August 2009, starting with new business in workers’ compensation. "We were surprisingly pleased with the amount and quality of our data, but due to resource limitations on our end, it took us longer to have the model developed than we might have liked,"
Parks says.

The SaaS-based system uses structured policy or application data and automatically draws data from external sources, such as the Bureau of Labor Statistics. Based on the analysis, underwriters are presented with a pricing recommendation. How underwriting acts on that recommendation is captured by Society’s enterprise BI platform for traditional reporting analysis.

"If the model recommends a certain price point and the underwriter decides to go a different direction, we can analyze how that turned out and either fine-tune the model or educate our underwriters," says Parks.

A few months into the system, Society is on the path to achieving its objective of driving greater consistency into the underwriting process. "We didn’t want to roll out a hands-off model to make an up or down decision, but it’s given the underwriters a point of reference for pricing," Parks notes. "Particularly on reviews of unprofitable books, we’ve found the results pretty striking—situations where if we had turned the clock back and applied the pricing the model had recommended, our results would have broken even."

Predictive analytics also is becoming a business necessity, continues Parks. "We believe at some point, our competitors will be using this technology, and if we don’t, we will be at a competitive disadvantage," he points out. "Particularly in the small account market, accounts can look very similar but be substantially different. Putting BI in the hands of underwriters to help them find those differences is the key to gaining a market advantage."

Society plans to roll out the system to other lines of business and processes. For instance, premium audit could develop a model to determine what types of accounts would be most likely to generate additional premium. "We continue to build up our base of data points and look for more areas of correlation," says Parks.

Society sees the potential, as well, to better manage its independent agency force. "We think there’s work that can be done to analyze our agents more effectively and determine how different agencies, or different tenures of agencies, perform," Parks says. "Those are the types of analytics marketing people do now through traditional reporting, but we think there is upside to being more precise on how different data points impact each other."

Agency management is a natural fit for analytics, asserts Ferguson. "BI helps choose the right brokers to do business with and get a good channel going," he says. "You need to direct your efforts to brokers who are more profitable."

Producer management was a key objective of Zurich, which has put BI in the hands of the users through SAS’s business analytics platform. "Once we’ve built a predictive model, we look at a territory and identify segments of a book we want to go after, then work with our direct market agents to target that business," says Joel Appelbaum, chief analytics officer of Zurich North America’s commercial programs and direct markets. "We’ve optimized our marketing by prioritizing it based on analytics."

Underwriters use the system for individual risk decision support. "We used to come up with an average class rate and some characteristics that would differentiate it. Now, we’re looking at future projected profitability based on inferential statistics," says Appelbaum, adding analytics has helped Zurich trim underwriting expense.

"We can save money on ordering MVRs by predicting which accounts are likely to drive in more congested areas, based on neural network patterns we developed," he says.

In addition, analyses can determine whether intervention or more service is needed to prevent problems from occurring. For instance, because Zurich has identified a correlation between an account relocating its business and a spike in claims, loss control engineers can use the system to prioritize those accounts for reinspection.

Although users are provided the tool, they ultimately have the control—and responsibility—of whether to use the intelligence it provides or not. "We create the model and provide the KPIs to our staff, but you have to give your staff the latitude to run the business," Appelbaum says.

Zurich estimates a savings of $5 million annually from incorporating predictive analytics into its overall BI effort, gained from both expense and loss reduction. "That’s not a lot—roughly one percent of our loss ratio—but every little bit counts in this game, and we think as we implement more, we’ll see additional savings," Appelbaum says. "I also look how profitable we are in the pit of a soft market, which shows our discipline to pricing and underwriting that is supported by analytics."

In today’s economy, "a 101 [combined ratio] is not what it used to be," Carnahan says. "In order to generate a profit, insurers must find ways for business users to leverage BI and truly improve performance. And particularly today, when the insurance market is shrinking, it is imperative to generate efficiencies to get a greater return, improve underwriting results, retain and sell more to customers, and get new customers."

Olympus Insurance has minimal IT staff, outsourcing its policy and claims administration systems and business processes. The company relies on its BI platform for performance management and to provide an important control process between its own staff and its outsourcers.

"We want to focus our IT resources on product development and financial management rather than running the ad hoc queries we need," says Brad Burton, vice president of underwriting at Olympus. "We wanted a centralized source that both we and our outsourcers could provide data to and that would deliver information to us on our desktop on a daily basis. We also wanted a resource that would be available to managers whether they’re in an office, at home, or on the road and that would provide comprehensive and consolidated intelligence."

Olympus uses a Web-based BI application from iPartners with access provided to everyone who is an operational manager. Olympus establishes the KPIs for performance management on a corporate level, and users are provided the flexibility to conduct analysis within those parameters. For example, users can’t change the dashboard and KPIs that are measured, but they can drill down those measurements when problems arise rather than asking IT for more data.

Burton illustrates this with a recent example from underwriting. "One of the marketing managers noticed some underwriting transactions that seemed to be increasing in her area, including an increase in cancellations for accounts that didn’t meet underwriting criteria," he says. "Rather than contacting IT for a query, waiting for that, then finding it didn’t have the information she needed, she could drill down into the cancellations that were made, why they were processed, and what territory they were for," he says.

"With the iPartners system, the answer to this question could be figured out much more quickly than it otherwise would have. It’s a more comprehensive way to view the data, with many more data elements than you would get in a report," Burton adds.

Olympus plans to add predictive analytics to its BI solution, a development that will happen "sooner rather than later," according to Burton.

"Our current environment allows us to gather and dissect information. We built the ‘slicing and dicing’ part upfront. The next piece is to take our book of business, internally analyze all these data elements iPartners has supplied us, and create an analytics model," Burton says. "We want to get that information in the hands of the sales force in real time to target business that will be profitable for us."

Despite the proliferation of new tools, BI relies first and foremost on good data. "One of the challenges of setting up [an analytics tool] is proper treatment of missing or poor data," says Appelbaum.

"There are many impediments to successful BI," says Carnahan. "First is often insurers don’t have an enterprise data model. They don’t have consistent data across different applications, so they can’t pull data from across applications in a consistent way. That includes basic problems such as the policy admin system using a ‘policy issued date’ and the claims admin system using ‘age of policy’ to reflect the same information."

And despite the apparent wealth of data at a typical insurance company, many insurers remain information poor. "You will find the situation where companies don’t fully capture risk information until a claim happens," Ferguson says. "So, when it comes to BI systems, an insurance company often is lightweight on insured risk data. And in addition, if it gets a lot of its business via brokers, it is lightweight on customer data."

"Other challenges to creating value include difficulty making changes to the core systems in order to capture the right data elements. Those systems often lack reporting capabilities, they lack ETL capacities, and insurers themselves often lack information distribution capabilities," Carnahan says. "Additionally, it’s very easy for users to get tricked by a false sense of precision with predictive analytics. You have to remember with predictive analytics, it’s a best guess—just because a claim scores high on a fraud scale does not mean that claim is actually fraudulent."

Unstructured data, where text mining approaches are needed to extract gems of intelligence from mountains of potential information, continues to present challenges for insurers, too. "Insurers have to invest in data integration tools and some kind of text mining in order to get the data out about insured risk, which would give them a lot more knowledge about claims patterns arising from certain types of risks," Ferguson says.

But for insurers that can solve these data challenges and put BI in the hands of business users, payback awaits.