Rethinking Data Warehousing
THIS MONTH'S EXPERTS
MARTY SOLOMON
Business Systems Architect, CIGNA (Philadelphia).
MIKE SCHROECK
Partner and Global Leader, iAnalytics Practice, PwC Consulting (New York).
SHARON SIBIGTROTH
Managing Director, Strategic Data Management, AXA Client Solutions (New York).
FRANK TEKLITZ
Director of Strategic Marketing, Business Intelligence, Sybase (Dublin, CA).
KEVIN YORGENSEN
Assistant VP, Information Services, Property-Casualty IT, The Hartford (Hartford).
RON HANKISON
President and CEO, Xbridge Systems
----------------------
Q: The conventional wisdom is that data warehousing generally hasn't been successful in insurance. In what ways do you agree or disagree?
A: Mike Schroeck, PwC Consulting: Many insurance companies have built very large and useful data warehouses (DW). Some have not been as successful because the work was performed primarily by the IT organization with minimal involvement and sponsorship from the business. More recently, business executives are demanding better and more timely information and, therefore, the investments these companies have made in building the DW infrastructure will enable them to deliver on these requests in a much shorter timeframe. As a result, many insurance companies who have invested in DW will only now start realizing the tremendous value to be derived from these investments.
Another reason that some insurance companies have not succeeded with DW is that they have approached this through separate, non-integrated efforts within multiple functions (e.g., marketing, finance, agency, claims, etc.). The result is that they have ended up with numerous data marts that lack consistent, accurate data and cannot be easily integrated.
A: Marty Solomon, CIGNA: I would disagree. Based on my experience and those of my colleagues in other industries, I would say that DW initiatives are no more or less likely to succeed in insurance than anywhere else. Perhaps success or failure is more difficult to quantify or measure in insurance, since the inherent data properties are generally more complex than in other industries, such as sales and manufacturing. As a result, the insurance DW will contain more complex extract, transform and load (ETL) processing and more data nuances than most, so the cost and effort of designing, constructing and implementing those warehouses tends to be much higher. Consequently the expectations, are often raised and a higher standard is set for declaring the given project a "success."
A: Sharon Sibigtroth, AXA Client Services: There have been successes and failures. Failure factors include lack of consistent commitment and recognition that DW is a core component of IT infrastructure, not the latest hot project; underestimation of data issues that will be encountered; lack of business skills and knowledge of how to analyze the data and identification of what business problems are to be solved; multiple business users disagreeing on currency of data, granularity, etc., tending to eat up analysis time; time spent on "tool battles" with vendors and integrators; and too little investment in measuring successful utilization.
A: Kevin Yorgensen, The Hartford: The success or failure of DW initiatives is situational, in any industry. In many cases, even within one company, there are various perspectives regarding the value that a DW initiative adds to the business. Most often, this depends upon the breadth of business sponsorship and the business functions that the warehouse supports.
A: Frank Teklitz, Sybase Business Intelligence: The difficulty factors that all DW in insurance have in common are cost and risk. They are huge, often measuring in hundreds of terabytes of data. Costs of building and managing very large data warehouses (VLDW) can be daunting, and the traditional database players are not delivering cost-effective, functional solutions. Insurance companies need VLDW solutions that deliver scalability, speed, flexibility and a much lower total cost of ownership. By looking critically at vendor products and measuring their capabilities in these four key areas, companies in the insurance industry can reap the business value of a DW.
A: Ron Hankison, Xbridge Systems: By its nature, the insurance industry magnifies an inherent limitation of the DW, which is designed to make all current operational data available to users and to retain historical snapshots at intervals. Because the data sources comprise shrink-wrapped plus in-house developed applications, cleanup and standardization are essential, as no data standards exist in the source data. In the insurance industry this problem is magnified as there are so many data sources, not all of which exists in a machine-readable form. This means that the data is incomplete, negating the very "completeness" goal of the DW.
Q: What are the success factors that contribute to successful data warehouse initiatives?
A: Schroeck, PwC Consulting: We feel that the number one success factor contributing to successful DW initiatives is the active involvement and sponsorship from the business in order that the DW can be designed to deliver real value in addressing important business opportunities.
Another important consideration for insurance companies is using an iterative approach to DW, starting with a single DW Application. As part of this approach, they should have an overall vision (data standards, technology standards, processes, etc.) that will enable them to extend these applications across the enterprise.
A: Solomon, CIGNA: The clarity of aims or a specific set of key performance indicators and set of reports is essential to success. These requirements must stem from the highest levels of the organization in order to have complete corporate sponsorship. There should be full knowledge of what questions are to be answered by the DW, exactly what is to be measured, classification and business ownership of data, how it is to be reported on and how it is to be delivered. Clarity of requirements, accompanied with the requisite high level analysis and preliminary design efforts, will precipitate a good assessment of the total cost of ownership. It is also critical to manage expectations with regard to "maintenance" costs, since a successful DW implementation will attract previously uninterested users, thereby potentially increasing (and invalidating) earlier cost predictions.
A: Sibigtroth, AXA Client Services: Criteria include an enterprise strategy supported by senior management with the commitment of resources across multiple years. You don't have to take a "big-bang" approach with a multi-year project plan to "do it all," but you do need to develop a vision; define a strategic architecture at the enterprise level; define standards around data, process and tools; and identify DW as an integral part of the answer in support of business objectives.
In addition, a plan to include adding to/enhancing the data in the warehouse should conform to and align with project plans around new product or product enhancements so that the data is available to analysis as soon as possible following introduction of new products. There should be a concept of "end-to-end processing" that not only includes product development and enhancement to administrative systems, but also the following-through of reflecting that data in the warehouse for quick analysis to measure impact.
A: Yorgensen, The Hartford: DW can provide the single source of integrated information necessary to support significant process integration across functional areas within an enterprise. While this cross-functional integration can drive reduced costs, more importantly it can generate increased revenue through faster time to market and enhanced customer experiences. Attaining this value, however, requires unwavering senior-level corporate commitment to the cultural and organizational changes required when data is managed as a corporate asset.
A: Teklitz, Sybase Business Intelligence: This question talks to the cost of information, such as "How much will it cost me to manage one year of claims records?" It is also important to ask, "What is the value of analyzing one year of claims data, especially if your analysts need five years of data?" By understanding the value of the information, which is invariably astronomical in terms of ROI, all these other issues become understandable, manageable and often moot.
A: Hankison, Xbridge: For DW and other decision-support systems to be successful, the system and the data must be maintained, and the capacity and performance of the DW must be managed. Perhaps most importantly, the data must be kept current and consistent. You must determine if and when you will process updates, and if and how to process partial updates. These dependencies in update processing can get quite complex. Fortunately, the intelligent use of the data extraction, cleaning and loading tools, and the information catalogs can greatly ease the burden.
Anthony O'Donnell has covered technology in the insurance industry since 2000, when he joined the editorial staff of Insurance & Technology. As an editor and reporter for I&T and the InformationWeek Financial Services of TechWeb he has written on all areas of information ... View Full Bio
Tweet This