The day I started working on this article was the day I received the fifth privacy notice from my insurance company. (No kidding-I have five different policies with the same company, and each one apparently generated its own notice.) We can all thank Congressmen Phil Gramm, Jim Leach, and Tom Bliley for sponsoring the Financial Services Modernization Act that required these notices (the "GLB" act). There is other legislation for insurers in the life and health arenas to contend with-namely, the Health Insurance Portability and Accountability Act (HIPAA; see August TD). And can you say "Patients' Bill of Rights?"
So one thing is certain-while the GLB-mandated notices have all been mailed, the fallout from privacy issues is just beginning. Insureds will be getting riled up about data privacy, and that's a concern for an industry that makes its living from information. In today's climate, how can insurers mine their customer data to do what they need to do to stay in business?
What Can You Know?
First of all, let's establish some context-we're talking about doing business in the United States only. Granted, insurance is a global industry, but there simply isn't the space available in a year's worth of issues to cover what you need to do to ensure compliance with the ever-changing privacy standards of every country. Suffice it to say you'll have to comply with the policies of the toughest host country, and it's up to you and your legal department to figure out what those are.
The good news:You're still allowed to use customer information to conduct the business of insurance. (See the sidebar, "The Very Basics of Privacy Law.") You can gather customer information from your own application data, third party sources, and transactional experience, then mine those data to your heart's content. "In terms of analysis and data mining, most of the companies have not expressed any concerns or reservations to us in terms of privacy," said Ron Exler, director of research at Robert Frances Group, an IT business advisory service. "They believe that under the regulations as they exist that they have a right to do that. It's part of the benefit to them and the consumer to bring some of that information together and know the customer better."
In short, privacy issues primarily address action, not analysis. In that action, however, there is a potential concern. Consumers can "opt out" of information disclosure to unrelated third parties. On the face of it, that doesn't seem to be a concern for insurers, which generally use direct sales, captive agents, or contractually related third party brokers to sell and distribute their products. But, paradoxically, it could become a concern as the very combination of insurance and banking facilitated by GLB may create privacy problems for insurers looking to take advantage of an expanded financial services market.
In fact, the Robert Frances Group reports that "conglomerates offering a variety of related but distinct products and services spanning financial, insurance, banking, and brokerage services may be at an especially high risk for privacy invasion. IT executives managing such organizations should establish methods that ensure cross-selling and prospecting techniques are explicitly pre-approved by customers, or customers are allowed to opt out." Serious stuff, indeed.
And even if insurers are within their rights to use customer information for their own sales efforts, refusing to honor customers' "opt out" requests is foolish-akin to ringing the doorbell of a house with a "No Solicitors" sign even if local laws allow you to. It's a balance of personalizing service while not being pushy, and when it comes to successful selling, consumers' perception of the law is more important than the law itself.
What Should You Know?
It's downright impractical to exclude customer information from your data mining efforts, because data mining works best when you start with the most granular level of detail. "To have confidence in the aggregate, it needs to be built on solid detail," said Paul Theriault, Pinpoint Solutions' senior VP of marketing. The more customer information that's excluded or the more summary information that is used, the less confidence there is in the results. Data mining is about using detailed information you have on your customers to draw aggregate conclusions, and then reapplying those conclusions to individual policyholders and prospects. According to Theriault, it's about focusing on a specific "pain point." Where are you losing money, where can you make more, and how much are you willing to spend to address each issue?
The result of effective data mining is the ability to discriminate in the positive sense of the word-to use precision pricing and coverage fine-tuning to accurately insure risk, to predict what individual customers need and will want to buy, and to know what delivery channel each customer prefers to deal with. It can also help you address those pain points in unconventional ways. Perhaps analyzing a money-losing regional book of personal auto policies will reveal not that pricing is too low, but that repair rates in that area of the country are too high, which could lead you to negotiate preferred contracts with repair shops.
At the very least, it gives you the opportunity to show some financial return from that dusty data warehouse you spent two years implementing. (Assuming it's working. See "Promises, Promises" in the July TD.)
What Do You Know?
Successful data mining-being able to spot the trends needed not only for analyzing what has happened, but for predicting what will happen-starts, by definition, with the data themselves. And far from having too much information on customers, insurers in general have too little. Sure, it's easy to capture application information, third party financial data, and claims detail, but those statistics are only part of who your customers are. There's a wealth of information that you're probably not getting-things your agents know but don't tell you, information your call center hears but has no space on the form to record, and comments entered by customers on their bills or through Web site feedback, just to name a few.
"There are a lot of data that companies have that never gets codified," said Patricia Saporito, senior insurance industry consultant at Teradata. "It's all about communication-capturing all of the customer information regardless of how or where it's provided. If you write through independent agents, you should develop creative ways of communicating with the customer and getting feedback. You should not only put incentives in place to solicit information from the agents, but also use the call center. Every time the company has an opportunity to talk to a customer, it should be obtaining information."
With the data collected, there are two common models used to mine them: access the warehouse directly, or load selected data into a datastore dedicated to the mining operation. Loading a separate datastore is the more common model, because the disparate systems that hold data at most insurers are designed for transaction, not analysis. Even if the mining tool and your warehouse could work together directly, it still may pay to separate the logical data used by data mining from the physical data used to support your operation. Regardless, the system should provide a database architecture that can support low-level, multidimensional analysis. That means common OLAP (OnLine Analytical Processing) analytics, designed to perform predetermined queries and generate scheduled reports, just don't cut it.
"It you have 40 rating variables, and each can have 10 different values, that's 40 to the 10th power," Theriault explained. "No OLAP solution will support that. So at some point, your solution will demand an architecture that will support ROLAP [Relational OLAP]."
ROLAP allows you to create multidimensional views on the fly and to progress toward iterative analysis. "You gain incremental insight during the first pass through the data, and then you use that insight to synthesize something different and make another pass through it," said Eric Rogge, director of product management at WhiteLight. "Being able to perform multiple iterations this way allows you to more quickly arrive at a conclusion that can net you incremental profit gains."
What Will You Know?
Ready to start violating your customers' privacy for fun and profit? As with most technology decisions, you can try to build a system yourself, assemble components, or buy a prearchitected solution. If you decide to buy a complete data mining system, there are many vendors standing by to answer your call, and the following are a few who specialize in insurance. A vendor with insurance expertise should bring you prearchitected data models and insurance-specific analytic applications to speed installation time and increase system effectiveness. At the very least, it should mean that your first meeting with the vendor doesn't start with, "Tell me again what an actuary does?"
And as you build a system or evaluate vendors, you should keep the privacy issues discussed in mind. Most vendors toe the privacy-is-an-issue-of-action-not-analysis line. When it comes to selecting or building a data mining installation for an enterprisewide solution, as always, due diligence applies-understand the issues and make sure your system architects do, too.
The Very Basics of Privacy Law
Thanks in part to the Gramm-Leach-Bliley Financial Services Modernization Act (GLB) and the Health Insurance Portability and Accountability Act (HIPAA), privacy issues have become a big issue in 2001. Because the former has had more time to mature and affects every insurer (and because HIPAA provisions are not mandated for compliance until 2003), here's a brief rundown of key GLB privacy provisions.
What GLB does:
- Protects personally identifying information that is not publicly available ("Nonpublic Personal Information");
- Restricts information sharing with nonaffiliated third parties;
- Prohibits sharing of account numbers or access codes with nonaffiliated parties for marketing purposes;
- Requires initial and annual notices to customers describing privacy practices as well as any "opt out" rights
policyholders may have based on the insurer's disclosure practices;
- Allows insurers to use and share information with affiliates and with nonaffiliates to complete transactions
requested by the insured, to maintain the insured's account, to prevent fraud, or to comply with other laws
Data Mining and Privacy
Even if insurers are on the right side of the privacy law in their data mining operations, there are some privacy safeguards they should follow as a matter of good business practice, and as a hedge against future restrictions. Practice number one is "anonymizing" customer information. That is, while it is important to know as much as you can about a customer, until you need to take action on the information, you don't actually need to know who that customer is. Leave that to the customer contact system.
The second practice involves protecting data from inadvertent sharing that violates privacy law. "Data mining results can hold individual records, and most good [data mining] tools can be accessed over the Internet or be made to stand alone," said Kevin Pledge, lead consultant at Insight Decision Solutions, which provides analytical tools for life and health insurers. "You can take a slice of the data, take it away from the server. Therefore, you need to have good controls in place."
Jargon Alert
The term "data mining" is used throughout this article. However, equally as often in the industry, you'll run across the more impressive sounding "business intelligence." The difference? Business intelligence is the broad category of software, systems, and business processes that helps insurers better understand their customers and their business. Data mining includes the technologies and techniques to find and analyze data, and is a subset of business intelligence. To further confuse things, data mining is often used interchangeably with "knowledge discovery."
Click Here
Insight Decision Solutions: www.insightdecision.com
Thazar Solutions Corporation: www.thazar.com
Teradata: www.teradata.com
Robert Frances Group: www.rfgonline.com
Pinpoint Solutions: www.pinpnt.com
WhiteLight: www.whitelight.com
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.