If we measure it, we can manage it, and then we can start to make decisions to do it better. So says Gareth Morfill, Boston-based senior analyst at Ovum, the London-headquartered analyst and consulting company, in summarizing the goals of business intelligence (BI). But how can insurers best measure their performanceboth at the department level and across the enterpriseand who should do it?
The data for measurement is tantalizingly within reach, with terabytes of information maintained in data warehouses, data marts, operational data stores; that is, within the domain of the IT community. But its the business users who need to make the business decisions. So traditionally, users have had to beg, cajole, and even threaten their IT counterparts to create the reports they neededand then to continually modify those reports as those needs changed.
When I came on board five years ago, we had five individuals who spent a better part of their day responding to internal and external client requests for information, says Jim Klotz, CIO at PMA Insurance Group (Blue Bell, Penn.), a property/casualty insurer specializing in workers compensation and disability income insurance. They were using older tools, running ad hoc reports, doing queries with QMF [Query Management Facility, an IBM language for user interaction with DB2].
The Changing Big Picture
Enter business intelligence applications, the primary design of which has been to bring together data from the disparate sources at any insurer and allow the results to be analyzed, sliced and diced, and presented in any number of ways. But as these applications have gained traction in IT departments and power-user communities, some insurers are extending the power of these tools to more users to reduce the demands on IT, and still other insurers are asking, Why arent we?
For starters, the tools havent always been as good as they are today. Historically, the user interface was a challenge, says Tom Chesbrough, executive vice president at Thazar, a Skywire Software company, which provides insurance-vertical analytical applications. There was all this talk about pushing information to the users, but it was cumbersome and complex. Unless they were power users, they werent interested. With better interfaces and the explosion of corporate portals, this has begun to change. Users can customize [portal] dashboards, and they can view information in a variety of ways. Sales sees information in terms of production, underwriting sees it in terms of pricing, and so on.
Equally important, cultural issuesthe basic fear users will screw things uphave worked to slow the extension of BI tools past IT. Ive seen IT limit business users access to the data because of lack of user training in the reporting tool, limitations of the reporting tool itself, lack of understanding of the data, lack of user interest in defining their own reports, [and] IT control issues, explains Patricia L. Saporito, industry partner in the insurance & healthcare space at Teradata, a division of NCR.
The tools havent added more functionality; companies simply havent extended the function of what was already there, Morfill says. There is a complexity associated with deploying BI technology, but weve matured in our understanding of it to the point where we are now in a position to deploy it to a larger audience.
Understanding that audience is the first thing insurers looking to extend the reach of their BI tools must do. Beyond the realm of IT, a whole spectrum of computer literacy exists at any insurerfrom the power user, to the computer savvy, to the technologically challenged. We are seeing a shift to users becoming more able to at least take a template of a report and turn it into the report they need, Morfill adds. Were starting to work toward a stage where organizations have certain individuals who can create a report from scratch. But were still several years from seeing the really high-end data mining and analytics becoming deployed as tools on every managers desktop.
Tales from the Trenches
The carriers that spoke about their BI solutions have all extended some level of functionality to different tiers of their user base and are looking for ways to do even more. PMA created a suite of Web-based reporting and analysis tools using Brio, implemented with the assistance of consulting firm Idea Integration at the insurer in 2000. Today, the tools are extended to about 2,000 users, including PMA staff and some clients, agents, and brokers.
Different components of the solution target different types of users: PMA Cinch is the extranet application designed for customers and business partners to obtain static reports and data analyses via the Web. Most business users within PMA have access to PMA Insight, which provides about 80 preformatted queries and reports but also allows for customization of format. Power users within IT, actuarial, and other business units use Brio Explorer to run preformatted reports, create custom reports and analyses, and do ad hoc querying of the companys Microsoft SQL Server datamart and its transactional datastores.
We originally focused on the power users and currently have about 50 in that category, says Joe Flynn, PMAs assistant vice president of IT. Then the Insight users became more widespread, giving people internal access to existing queries. Finally, we extended the external system, training those users on the viewer to access the reports. The functionality to these three groups was available from day one, but the deployment has grown over time.
The stratified deployment was done with the needs of PMAs user community in mind. We had considered offering all users ad hoc tools, but theyve told us they prefer to have [IT] develop a custom report and have menu selections, says Klotz. If they want to perform additional analysis, they prefer to download the data rather than using an ad hoc tool within the BI environment, and then use whatever tool theyre accustomed to, such as Excel.
La Suisse, a Swiss-based life and property insurer, installed DecisionSite by Spotfire to track sales against corporate objectives and evaluate how products are selling across customers and geographical regions. This allows La Suisse to keep track of the status of renewals of its policies and the time it takes to process life policies. Additional areas of analysis currently are being developed.
Spotfire is solely client-server based and requires installation of the user interface at each client PC. At La Suisse, Spotfire runs on Windows NT. Installation began in January 2002 and was completed in four weeks. The system queries the insurers AS/400-system via Access databases, then transforms and analyzes the retrieved data, delivering the results in reports that are predefined by La Suisses sales department.
Currently about 100 users, or 10 percent of La Suisses total workforce, have the Spotfire tool. This relatively high percentage includes everyone from users [who know] programming to people who have only a basic understanding of Excel, says Gabriel Fuchs, senior manager in the insurers sales and marketing department, who is responsible for the DecisionSite installation. Also, if users decide they dont want, for example, the predefined pie charts and bar charts, but rather plotter charts or Excel spreadsheets, they can change the report format. If they have other datafor example, in Excelthey can integrate it simply by cutting and pasting that data into DecisionSite, Fuchs explains.
American Home Assurance, a division of AIG (New York), installed ProfitCube by Pinpoint Solutions for its warranty policy business. We underwrite millions of contracts per year, so we needed a database management system that would allow us to have a better understanding of the contracts in force, the expired contracts, as well as the expense by type of contract in the database, says Matt Frankel, vice president of the warranty division.
AIG has extended the Pinpoint system primarily to its power users, including the account management and actuarial team as well as underwriting managers. This decision reflects the fact AIG looks to the solution as a way to analyze its book of business for long-term strategic decision-making, rather than day-to-day tactical information.
Frankel explains: There are 33 different parameters, and we can query any or all of them to understand the business better. For example, if I wanted to take a look at color-television loss rates over the past 27 months in a certain 5-digit ZIP Code in the state of Georgia, I can do that by just a few clicks of the mouse. So it allows a nontechnical user to get an answer to a simple or complex question easily, without needing help from IT. With millions of contracts in force, [this] was not something we were able to do before installing ProfitCube. Configuration of pa-rameters, loading the historical data, and properly matching it to ProfitCube typically falls to AIGs IT department.
More Users = More IT Demands?
Extending the functionality of BI tools to the user community is done in part to reduce the strain on IT resources. However, any distributed deployment brings with it the risk of new issues that didnt exist under centralized control. Fortunately, sources report extending BI does result in net time saved for IT.
Whenever theres a new application, theres a new workload, but the tools are robust enough, particularly being Web based now, so that a lot of that previous overhead has gone. You typically dont have to go around to hundreds of desks to install a client on them, says Morfill. Also, most of the tools have a very intuitive way of working, but if even thats too difficult, you might integrate those tools with a portal, allowing users simply to click on the report and type what they want.
We identified wed need to have some sort of help-desk assistance, says Klotz, but the good news is the tool is easy enough so that were not spending a lot of time [on help-desk issues]. Most of the questions come from external users. Whenever you extend a tool outside the campus, there will be questions. If the application is having a problem, the clients will call you first, not their own IT shop, so you should be prepared to offer support. Because of that, we keep what we extend to the external users easy to use, and we deployed Citrix to minimize network access problems.
In fact, says Flynn, critical issues of BI still involve data quality, rather than the use of the BI tool. As more people are exposed to the data, clean data becomes increasingly important, as does defining the data. The problem of two people looking at reports, but seeing different results, is the result of extracting data differently and not due to the BI tool. Good data dictionaries, good ETL tools, good data definitions, those are the most important.
BI Lite?
The most basic definition of business intelligence is simply information that allows people to make better decisions. But this literal definition always also has carried a strategic implicationperforming BI meant mining reams of historical data, uncovering trends, and performing analytics on the enterprise level.
Still, thats not stopping all flavors of vendors from getting in the act: Express an interest in business intelligence, and every vendor from A to Z will be dropping you a line saying they do it. There is some basis to these claims, however, since more and more vendors of processing systems have added reporting functionality to their offerings in recent monthsin essence, offering BI Lite.
We are seeing new additions [to transactional systems], says Morfill. To some extent, these systems have been collecting this data forever, so its a natural extension of functionality. And many of the [transactional] tools are partnering with BI software products.
He points to document management vendor Filenet as an example. Filenet is making use of Microsoft SQL Server, with its range of data warehousing capabilities, to provide a variety of reports you can drill up and drill down into, to present information as graphs, to provide template reports showing how long documents have been in each stage of the process and what the exceptions were, and so on. An example from the claims management space is Castek, which has added management reports to its Web-delivered, ASP-based ClaimsPath tool within the past 18 months.
Kingsway General Insurance Com-pany (Mississauga, Ontario), which began using ClaimsPath for management of claims in October 2002, has found these added reports particularly valuable. We are able to generate productivity reports anywhere from our adjusters to our service providers, says Matthew Turack, special projects supervisor at Kingsway. The system allows us to track our vendors and their productivity to meet our business rules, such as car rental agencies responding to us within 30 minutes. With these reports, we can determine which vendors do not meet our standards of customer service. Similar reports track percentage of assignments to our preferred vendors, and based on the results of such reports, we can assess the current business relationships with these service providers.
Policy data from Kingsway is sent on a nightly batch to Castek (real-time replication also is an option but has not been implemented at Kingsway) and is used to populate fields on first notices of loss and to confirm coverage data. Claims data is maintained in a database at Castek and updated in real time as adjusters process claims transactions using ClaimsPath. Management reports run against the claims database are generated on demand based on user-entered parameters and are delivered as HTML in real time. Code modifications of existing reports currently are done by Castek. Reports also can be converted into Excel spreadsheets and manipulated, Turack adds.
The upside to added-on analytics is they are coming from processing and management systems insurers already have in place or are going to purchase. In fact, according to research firm IDC, about one-third of the intelligence businesses, including insurers, obtain comes from processing systems, while two-thirds comes from dedicated analytic and reporting systems.
I would expect that ratio to remain in this range, says Dan Vesset, research manager in IDCs data warehousing and information access program. The two should be used in conjunction with each other, however, because with an analytic data store, you can integrate data and analyze historical trends, whereas on a production system, youre just doing some very specific reporting on a narrow data set.
Processing systems also have a leg up on traditional warehouses when it comes to providing access to up-to-the- minute data. However, its unreasonable to expect the same level of analytics from tools designed principally to handle workflow or processing and those designed solely for BI. Primarily, expect little capacity to customize reports and queries, and try not to incorporate data external to the transactional information the system itself stores.
At some point, youll want to integrate those [add-on] analyses with other key data to enrich the information, says Morfill. If its only one or two key aspects youre looking to enrich it bysay, youre looking at your process flow, and youre looking at adding some customer informationthat might be only two or three fields. That could add a lot to the value of reports you can get out of that single system, but whether you can add it depends on the architecture of the system. If its doing the reporting off the operational database, rather than having a separate reporting database, then its a lot more work.
Insurers have these canned reports, but overall, business cant manage through canned reports because they dont allow the correlation or aggregation of, for instance, the application system with the financial system, says Neil Patil, Brios vice president of product marketing. For example, to evaluate whether rates have influenced an increase or reduction in claims by specific geography requires you to go across systems.
But is there any risk in using add-on analytics as is? According to some, there is. Because production databases arent tuned and optimized for querying, running analytics against them causes performance problems because 100 percent of the time those reports are hitting the production system, says Frank Gelbart, CEO of Appfluent, a provider of software designed to improve enterprise application performance.
Not so fast, say others. The way that we deal with [performance issues] is twofold, says Colin Smith, vice president of eClaims at Castek. First, the report modules are not core to the system, and the processing can be offloaded to secondary servers, such that the report processing does not directly impact the transaction processing servers. Second, much of the information either is precompiled or partially precompiled to speed up report processing, although the user can still compile reports from the base data. Also, the insurers that spoke about their BI installations did not report performance as an issue.
Therefore, the bottom line is if your users have reports from add-on analytics they like, and nobodys complaining about system performance, by all means keep using these reports. But consider replicating these reports as part of your overall BI strategy both to enhance the intelligence with additional data and to provide the resulting analyses to a wider user group.
And not to be overlooked, no tool can ever replace savvy business sense. Business intelligence tools provide insights into our business, but not answers, Morfill concludes. We might see a trend, such as on Mondays things take twice as long to process as on Fridays, but that doesnt tell us why. You cant go to the computer and say, Whats going on? Ultimately, even the best BI tools only provide information; its up to the users to identify patterns and try to infer what may be the causeor the opportunity.
BI Vendors
Appfluent Technology
Arlington, Va.
703-284-0800
www.appfluent.com
Brio
Santa Clara, Calif.
800-879-2746
www.brio.com
Castek
Toronto, Ontario
416-777-2550
www.castek.com
Guidewire Software
San Mateo, Calif.
650-357-9101
www.guidewire.com
Information Builders
New York, N.Y.
212-736-4433
www.informationbuilders.com
NCR/Teradata
Dayton, Ohio
937-445-5000
www.teradata.com
Pinpoint Solutions
Livingston, N.J.
973-716-0723
www.pinpnt.com
SAP
Newtown Square, Pa.
610-661-1000
www.sap.com
Spotfire
Somerville, Mass.
617-702-1600
www.spotfire.com
Thazar
Overland Park, Kans.
913-327-7881
www.thazar.com
FACT: Support for ad-hoc queries is the most important factor insurers consider in selecting a business intelligence product.
Source: Gartner Research, Research Note, October 30, 2002.
© 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.