Heavy Lifting

BI Gets High Marks

Highmark Blue Cross Blue Shield currently has a series of data warehouses that support different line functions: health-plan membership, claims and medical expense, and call center information. We link data across all the environments, but in terms of a purist definition, we dont have a single enterprise data warehouse yet, says Darren Macioce, the Pennsylvania health insurers vice president of healthcare informatics.

In addition to typical management reporting and financial reporting, High-mark uses these data environments to create detailed intelligence such as physician and hospital performance profiling and predictive analysis. Highmark has also used BI for proactive measures. For example, algorithms that analyze customer claims information identify individuals who may benefit from various clinical intervention and educational programs for diseases such as diabetes and coronary artery disease. That allows us to identify those individuals for early intervention, to help them better manage their health status andpossibly reduce their medical costs, Macioce says.

Over the past two years, the company has migrated its warehouses from DB2 and Oracle to an NCR Teradata environment. Thats not only allowed us to produce seven times the number of queries we could a year ago, but resolving performance issues has gotten us out of maintaining a bunch of data marts and cubes, and allowed us to be more creative in the types of analysis that we do and the issues we study, Macioce says. Cubes are multidimensional databases that hold summarized data more like a 3-D spreadsheet rather than a relational database.

The company also utilizes a variety of analytic tools as point solutions in various departments. We have big investments in a variety of SAS products for heavy lifting. We have a good bit of Webfocus for customer reporting, as well as some smaller applications supported by Cognos, Macioce explains. Creating the informatics department three years ago was designed to help consolidate Highmarks BI approach. Were trying to decide whether any one of these tools will do the job, although were getting a lot of value out of what we have today.

Many warehouse-based BI initiatives fail because companies lose control of the scope of the project. Thats one reason ACUITY, an 11-state regional property/casualty carrier headquartered in Sheboygan, Wisc., has chosen to deploy analytic applications on a department-by-department basis to provide intermittent deliverables but build those applications on an enterprise warehouse architecture.

We started out several years ago with a small, but very defined, pilot, working only with personal lines auto, explains Neal Ruffalo, vice president of enterprise technology. For the pilot, the insurer partnered with IBM Global Services to define user requirements and select appropriate technology, a process that heavily involved both IT and business staff to define not only the ultimate analytic goals, but also the data elements, data models, and metadata.

That was crucial, explains David Jablonski, director of information systems. Previously, we had produced only the paper reports common to most insurers. To have users think about not only the data they wanted but how to manipulate the data to do analysis without coming to IT generated a lot of interest.

With the exception of its financial system, all of ACUITYs administration systems were developed in-house and use three centralized data repositories, an architecture that has aided its warehouse development. We were able to write a COBOL program for the extract, rather than bringing in an ETL [extract, transform, and load] tool, Jablonski says. We have data down to the coverage code and cause of loss levels, which is needed by actuarial, but have also developed summary tables for other areas as weve gone forward. Strongupfront edits have also helped keep the data clean over time.

ACUITY evaluated three query toolsCognos, Brio, and Business Objectsover a three-month period and selected Business Objects in the end. Currently, the system is licensed to support power users throughout sales, accounting, actuarial, strategic information, IT, and marketing. Eventually, we would envision empowering everyone in the organization with some access, but currently access is at the strategic decision-making level, says Ruffalo.

The system has had an impact on personal lines, the first area targeted. The company recently implemented a single-plan auto program that entirely replaced a portfolio of four separate plans. Tiered rating, including credit scoring, is critical to the rate adequacy and underwriting accuracy. We were able to calculate loss ratios in very fine gradients, to track the impact of credit score on profitability and set our pricing to a much greater degree of accuracy, says ACUITY president and CEO Ben Salzmann. We could instantly generate and fine-tune analysis, do modeling, various what ifs, and iterations to fine-tune the program. The speed and fluidity of the analysis easily saved us half a year of time to market while simultaneously giving us a great level of detail.

Broker relationships are the next area of the insurers BI focus. We can track average credit score by agency, or SIC code by agency, and then break that down by sales manager or line of business. I can get that in an Excel spreadsheet and manipulate it further, or send out a PDF of results to an agency, says Brian Benishek, regional sales manager.

If youre looking for dollars-and- cents ROI, we cant put our finger on it exactly, says Ruffalo, but its one of the reasons weve grown in double-digit percentages in personal lines while maintaining profitability.

Data (S)Mart Strategies

A data mart strategywhere smaller databases are deployed to address individual departments or tactical needscan be effective as long as an overriding architecture to connect the marts together is followed (see Data Warehousing: What Works? July 2001). Thats the method used by Pinnacol Assurance, the largest workers compensation insurer in Colorado. We have data marts in almost every area of the company, but we have no stovepipe marts, says Bob Pfahler, decision support services manager. The connection is a collection of shared dimension tables that link data marts and aggregate tables where necessary for performance.

Pfahler reports a two-fold advantage to a data mart approach. First, we can bring up new marts pretty quickly. Second, if youre completely going down the wrong road, you havent lost much.

Over half of Pinnacol employees use Brio to track their own performance against goals and the performance of their team, department, and company. We go right down to the desk level for claims, to the underwriting level for policy, and to the line level for bills, says Pfahler.

When bringing a new data mart on line, Pinnacol first includes only current transactional data before loading historical data, a process that speeds implementation. We can implement most subject areas in about 60 days. The backfill [of historic data] takes longer, about six months.

Business buy-in has been strongat Pinnacol. Every major data mart has come out of a strategic business plan, rather than IT dreaming what data marts we need to construct. There are no execs who have not participated in one or more data marts, says Pfahler. Thats important because the numbers are going to hurt some people and benefit others, and the people who are hurt by the numbers need to see that as an opportunity to improve.

Though no cost figures were available, Pfahler believes the project continues to justify itself over time. How I measure our ROI is what the company [management] is willing to spend on data marts for the next year. They never come to me and say, Tell me how much it returned to us. Instead, they come to us to be involved in policy pricing and issues of strategy.

Leaving No Legacy

Life insurers in particular are faced with long-term contracts and years of legacy data. With 155 years in business, Canada Life is no exception. In 2000, the insurer converted to a new policy administration system to address legacy performance issues, while at the same time looking for an analytic application to address BI shortcomings. Previously, if management wanted to see a type of report, they would go to IT, and the turnaround would be a month, says Michael Nowski, senior business consultant and manager.

The insurers first area of focus was producing experience studies that had previously been performed by the now-decommissioned legacy system. The choice was to write a new application or to bring in a platform that was flexible and scalable to address future analytic needs. Canada Life chose Insight Analyst by Insight Decision Solutions, Ontario. We have several admin systems, one for policies, one for reinsurance, and one for claims, Nowski says. All those feed to a staging area that in turn loads the Insight database.

The heavy lifting of data conversion occurred with the change to the admin system itself, so implementation issues of the Insight system were minimal. Currently, the insurer has converted about three-quarter-million policies to the new administration system. The biggest issue as you would expect was data quality, explains Nowski.

Principal users of the Insight system are the actuarial communities within the company as well as some business analysts. The main benefits to the Insight tool are the flexibility in reporting of results. Before, wed have canned reports generated off the mainframe. In the new world, the system is user defined, and everything is based on a cube.

Canada Life intends to expand its BI capabilities by interfacing its Siebel CRM system to the warehouse. Were planning to include data on our 6,000 producers, their results, and their relationship to the sales office, Nowski explains.

Better BI

These insurers results point to just some of the impact that BI can have. In underwriting, companies could be looking at aggregated exposure, looking at spatial analysis, trying to visualize exposures, says Pat Saporito, industry partner of insurance and healthcare at NCRs Teradata division. They should be looking at better profiling of their customers, looking at the right time and right products to cross sell based on customer behavior. They should be narrowing their marketing programs to spend less and get a higher response rate. They should look to negotiate better contracts with claims suppliers based on average costs and settlement times.

Also important, insurers should have a sound business reason for buying or building a BI solution before even considering an implementation. Regardless of industry or country, the first reaction by companies is to buy a tool, says Giga Information Groups Gile. Theyll get power users on board, and occasionally these companies will get IT involved. But no attempt is made to scope the project; no one establishes the requirements specifications; no one performs a readiness assessment; no ones concerned about testing or quality control. So what you end up with is a proof-of-concept application that when successful, companies try to extend to other departments, only to have it blow up because it wasnt designed to be scalable.

Another critical component to the choice of tool is the vendor that developed it. On one hand, insurers can go to a large, horizontally focused vendor that may offer better stability but lack domain expertise; on the other, to a niche vendor who might offer the inverse (see Choose Your Vendor Intelligently, p. 32).

Additionally, insurers must make sure the chosen tool is encompassing and powerful enough to answer more than initial business questions. A single unified view of your customer is critical, says Gile. Otherwise, what you end up with is a CRM system that affords one perspective of the customer, a marketing system that affords a second perspective, and an ERP system that delivers yet another. You need to take into account integrating these disparate data sets, how granular the data must be for relevant and meaningful analysis of these combined data, and whether aggregated, detail, or both types of data are needed. These are not easy, one-word-answer decisions to make. They require discipline and structure.

How Much History?

Few things present more problems to a warehouse-based BI project than historic data. Over time, fields may have been added, transactional data may have been lost, and data may just be plain dirty. (As the saying goes, Never underestimate the creativity of the user when faced with a data-entry edit.)

Therefore, some installation strategies start by going forward only. Yet eventually, users are bound to clamor for a bigger picture. Some insurers have told me when they did their financial analysis, it tended to be over 50-year cycles, Braunstein says. Think about the business. The objective is to collect money, invest it over an extended time, and hope that the payouts tend to be less than your income stream. And the business goes through cycles, whether P&C, life, or healthyou make profit in some years and not others, and you hope through the economics of large numbers over time you play it well.

Over the years weve seen both types of implementations, from going-forward models to long history conversions, says Gareth Morfill, senior analyst in the Wakefield, Mass., office ofthe London-headquartered consultancy Ovum. The usual reason for going with the well turn it on and start collecting history today is that someone has messed up the budget or doesnt realize how much disk space and how much processing they need. The decision really comes down to having a clear understanding of the type of analytics one is looking to perform as opposed to whether one wants the warehouse to become some sort of archive, rather than a BI solution. The further back in history you go, the more data formats there, the more complicated it becomes.

Choose Your Vendor Intelligently

Its no secret that the technology industry has been struggling in general. In the business intelligence application market, the consensus seems to be that every company is hurtingexcept for the company being asked the question.

This industry is near bankruptcy, says Tom Chesborough of Thazar, a BI company in Kansas City, Mo. There isnt a single one of us that financially was in a position to weather the storm out. He reports while business has started to pick up again, an infusion of investor capital was critical to keep his firm afloat. Pinpoint Solutions, Livingston, NJ, downsized its staff by 20 percent, primarily in sales and marketing, in late 2001 but maintains that sales are now looking up. Its been volatile on our end; the fourth quarter of 2001 after 9/11 is of course when our market dried up for that quarter, says Richard Baff, president. However, in the first quarter of 2002, we actually saw an increase in demand, and weve been rehiring and adding people.

Steven Bessellieu, president of North American operations for Sherwood International, says his companys full range of solutions from life to property/casualty to reinsurance has allowed us to take some of the waves more easily. The landscape is littered with vendors that havent proven out some of their offerings.

Bessellieu believes one of the problems that develop between insurer and vendor is an early strategic decision on whether to stay with a vanilla version of the implementation. Thats the way you drive a business development out of a package, he says. You are then going to depend on the maintenance from your vendor to provide you regular upgrades. Because you kept your system vanilla, they can more or less be easily applied, and you upgrade over time.

Many companies stray from that plan, though. They buy a package, maybe as a jump-start, but then they enhance it with their own solutions, interfaces, and requirements, and now that package is some percentage different than the original vendor solution, which makes implementing [new] vendor releases or enhancements either tough or impossible, says Bessellieu.

Millbrook, Bethlehem, Penn., didnt show at its booth at the IASA conference; however, the company reports this was not due to financial woes but, rather, due to a change in marketing strategy that eschews trade booths for more personal contact. Our revenue was up 85 percent in 2001, which was in turn up 65 percent over 2000, says company president and CEO Jack Plunkett, who adds that revenue has been flat through the first half of 2002 versus 2001.

What were seeing is that the market has matured in terms of the large, general-purpose software vendors: the Teradata, the Oracles on the database side; the Business Objects and Cognos on the Query side; and the Informatica and Ascential on the ETL side, says Gareth Morfill, senior analyst at Consulting firm Ovum. Theyve locked down the business for the most part. Some of the secondary players have had challenges as a result, and in the boutique [niche] vendor area, the economic downturn has hit them particularly hard because they dont have a wide spread of business or install base to take them through the downturn.

Vendors point to factors in addition to the state of the economy for any stagnation in the BI market. Weve had a tough time selling [our application for the insurance industry], says Pat Saporito, industry partner of insurance and healthcare at NCRs Teradata division. Insurers have been focused on CRM. Then they tried to fix their call centers and put in some kind of Siebel-type of application for contact management. Then they were focused on putting a Web front end on some of their existing applications.

She believes much of that work is well under way and the implementation of new policy admin systems will spur the BI market: Companies are becoming much more impatient about seeing their data.

BI Lite?

The promise of BI is to put the power of analytics in the hands of the user. The problem is the range of users who could benefit from these analytics varies substantially at any insurer, from the IT geek and power user to the business and casual user. Industry analysts report that as a result, tools need to become more approachable as they grow more powerful.

A number of solutions deliver their output into Excel, which is a familiar environment, and if users want to change the output, its a graphical, wizard-based interface, says Ovums Gareth Morfill. Weve also seen more Web delivery of information and a lot more on-screen help to step people through the process.

Thats essential to widespread adoption of BI, says Gigas Keith Gile. Power users constitute only about five percent of the end-user community, while the remaining 95 percent are the decision-makersmanagers, executives, partners, and customers.

Driving Bottom-Line Results With Business Intelligence

By Vinod Badami

Business intelligence is the solution at the heart of several enterprise objectives. It allows insurers to gain easy access to facts without having to access multiple application databases, perform sophisticated analysis to view the data across multiple dimensions, and analyze sales information to identify trends. It also allows carriers to analyze claims to reduce fraud, gain better visibility into enterprise expense structure, and develop new insight by combining data from multiple enterprise applications.

A leading health insurer offering multiple health plans in various states had customer membership, claims, and other related data distributed across multiple databases per health plan per state. Calculating key business measures such as customer and plan profitability, loss ratios, total claims and premiums, and fixed expenses by type of plan required access to multiple databases. A BI solution consolidated all customer, plan, claim, and expense information in a data warehouse thereby saving days in getting access to information. In addition, using this BI solution the insurer was able to compare plan profitability to determine which plans it should keep.

An emerging area of BI is the analytic applications that provide specialized intelligence about the enterprise. Insurers are discovering that while their CRM, ERP, and supply chain management initiatives have helped eliminate the spaghetti of operational systems and maintenance nightmare, the real value is obtained by implementing the analytical applications that use the data from these applications.

While BI applications deliver value internally to the organization, a new category that is gaining popularity is extranet applications. Helping customers and members access policy-specific data with self-service options to change coverage or service levels is especially attractive to Internet-savvy users who prefer to address their needs independently. BI extranets can also be used to help organizations that have purchased group benefit plans from insurance companies to better understand how their plan members are using the plan.

BI offers insurers many opportunities. The enterprise application binge of the 90s has helped create a platform of data that is ripe for harvesting. Insurance companies should utilize the following leading practices to avail themselves of the strategic value of BI:

1. Identify at a business process, functional, and enterprise level the strategic value BI can provide. Link these directly to the organizations strategic business objectives to develop a case for enterprise BI.

2. Based on the above, define specific BI initiatives prioritized by demand or level of pain. For each initiative develop a statement of payback and expected benefits.

3. Conduct a readiness assessment to determine the organizational capabilities to

successfully deliver each of the BI initiatives. Identify gaps at the infrastructure, resource, and delivery level. Use this information to develop an architecture blueprint.

4. Develop a roadmap showing how the gaps, resource plans, and delivery approaches for each initiative will be addressed.

5. Take a strategic enterprise level view and implement incrementally.

6. Develop a scope document for each initiative supported with a detailed project plan and deliver as per business expectations.

7. Focus on data quality, risk management, and cost effectiveness at each initiative.

Enterprise BI can no longer be restricted to a chosen few within an organization or considered just a good-to-have tool. To be successful with BI and get the value expected, one must approach BI with an enterprise view and identify individual initiatives to meet more immediate goals. BI can be a strategic weapon to drive significant benefits to the bottom line.

Vinod Badami is national practice director of business intelligence and data warehousing at RCG Information Technology (www.rcgit.com), Edison, NJ. He can be reached at vbadami@rcgit.com.