Unum Group, formerly UnumProvident, is the industry leader in disability income insurance and a top provider of voluntary benefits, life and long-term-care insurance products. The firm employs nearly 10,000 people worldwide and delivered $6.2 billion in total customer benefits in 2006.

Bob Dolmovich, vice president of business integration and data architecture, right, and Kyle Prescott, database administrator, share how foresight for future use of the Teradata Warehouse allows it to fill very different roles at Unum Group.

Today's Unum was forged in a pair of mergers. The first joined Provident and Paul Revere in 1997. The second, just two years later, merged the newly formed Provident Companies with Unum Corporation. This rapid convergence of three well-established organizations created an industry powerhouse on one hand and a thicket of management and integration challenges on the other, as each company had its own set of operational systems and its own approach to structuring business data.

Soon after the initial merger the company realized a data warehouse would be beneficial to support data rationalization and centralized reporting. "Our goal was to develop consistent performance metrics across the combined companies," recalls Bob Dolmovich, vice president of business integration and data architecture. "We needed to be able to track sales and a variety of other activity and profitability measures in order to understand more about our product performance and our distribution channels."

Building on a proven and scalable technology
In choosing the company's data warehouse platform, Dolmovich and his team made a strategic decision to build on proven technology. "We had enough business issues to deal with that we didn't want to be wrestling with technical issues too," he explains. "So we went looking for a scalable solution that may have been overkill for the data volumes that Provident and Paul Revere required at the time. We really wanted to ensure that the technology would meet our future performance needs. In hindsight, given the additional mergers that were just down the road, it was exactly the right decision."

In fact, the second merger—of Provident and Unum—was significantly larger in scale and raised the integration stakes commensurately. Ultimately, the scope of the combined companies drove a decision to blend all lines of business on an expanded Teradata Warehouse. In mid-2000 the firm launched a major platform upgrade and accelerated the efforts to assimilate and rationalize data from most of its 34 policy and claim administration systems, which included a wide range of mainframe, mid-range and distributed source systems.

The original data warehouse platform was a two-node system. Early applications for the larger data warehouse included a wide range of monthly performance analytics—everything from claim trends and premium trends and persistency—to daily sales reporting. Another priority was consolidating and integrating the many isolated and often redundant stores of customer data located in the various operational systems. To address this challenge, Unum implemented a Web services-based customer data integration hub, which in conjunction with the Teradata Warehouse provides customer data aggregation, standardization and quality control.

Evolving to active enterprise intelligence
As integration and rationalization efforts enabled by the data warehouse established a solid basis for enterprise reporting and performance analysis, Unum began experimenting with more frequent data loads and more active, operational business applications. "In many cases, after bringing policy and claim data into the warehouse for actuarial purposes, we were able to enhance the feeds or the timing of the feeds and ask ourselves what else we might do with the data," Dolmovich recalls. "Once you've done the initial hard work of accessing, transforming, loading and scrubbing your data, it's really easy to get extra value from it with relatively little extra effort."

Unum's more active applications vary widely in the company's business focus and in the types of workloads placed on the data warehouse. Two notable examples illustrate the diversity and scope of these applications:

	Individual disability valuation is the actuarial process of modeling aggregate policy values and the financial reserves required under law to meet projected claim obligations. Unum actuaries developed a more accurate and flexible process for managing nearly $12 billion in reserves by using the data warehouse to capture policy and claim change data from eight separate administration systems and recalculating reserve values daily. The calculation outputs alone occupy 750 million rows in the data warehouse. Business impacts include faster and simpler monthly closing, improved Sarbanes-Oxley controls and hard-dollar return on investment (ROI) that the company will describe only as 10 to 12 times its original projections.
	Interactive voice response (IVR) and Web self-service was a complex project aimed at improving the level of service offered to customers and brokers. Active data warehouse queries were used to support three separate types of interactions via three different customer and partner service systems: Intelligent call routing at Unum's customer service call center Self-service claim status inquiries on a customer-facing IVR system Underwriting status inquiries on a broker-facing Web portal
	Each of these applications resulted in a new active workload of tactical queries against the data warehouse. Additionally, hourly data loads from the various underwriting systems were implemented to provide near real-time visibility at the broker portal. In turn, these instances of active enterprise intelligence have significantly improved the level of service Unum provides to customers and partners, while substantially reducing the cost of delivery.

One data warehouse, multiple roles
As the tactical workloads associated with Unum's more active applications have grown in volume and importance, proactive workload management and monitoring have become a necessity. "Our current workloads run about 45 percent batch ETL [extract, transform and load] processes, 35 percent ad hoc and 25 percent tactical queries from the Web services and call center applications," explains Kyle Prescott, Teradata database administrator at Unum.

"We use Priority Scheduler to segment workloads and Teradata Manager to monitor the system," Prescott continues, "and we run canary queries. Our philosophy is to let people access the data, but to watch them closely. We're not using any denial-type software, but we do run lists of top-100 CPU users and top queries from a capacity perspective. We try to stay on top of that so we can protect those tactical queries. They're not a large part of the overall workload, but we think of them as our golden egg." This use of Priority Scheduler is especially important considering that the average query volume is greater than 10 million queries per month with peak volume as high as 18 million queries per month.

Unum's Teradata Warehouse now fills two very different roles for the company—one is historical and analytical, the other is active and operational. "It is certainly the pre-eminent business intelligence platform for our U.S. business, and it's the tool of choice for business analytics, pricing, valuation, reserve modeling and forecasting—all traditional roles," Dolmovich explains, "but it also has an increasingly important role in customer and partner service that it is able to fill because of its stability and 24x7 performance.

"And the Teradata Warehouse has given us something else: a fresh model for database management. Because we started this system from scratch just 10 years ago, with the benefit of prior experience and really good tools, we've done a much better job of clearly defining and controlling our data. It's a model for data management and governance practice that we're now taking out of the business intelligence space and back into our application development processes for transactional and operational systems." T

Behind the solution: Unum Group Database: Teradata Database V2R6.0.2 Server: 10-node Teradata 5450 Server Users: 2,500+ internal business users and thousands of additional broker and customer users via portal and call center applications DBAs: 3 (2 full time, 1 part time) Data Model: Third Normal Form Operating System: UNIX MP-RAS Storage: Spinning disk: 29TB. Customer data space: 7.9TB. Actual data: 3.4TB. Teradata Utilities: BTEQ, MultiLoad, FastLoad, Teradata Active System Management, Teradata Manager, Teradata Priority Scheduler and Teradata TPump Tools/Applications: Teradata Warehouse Miner and products from Business Objects, Cognos and Hyperion

Bill Tobey is a senior technology writer at Ford Sherman, a communication services agency.

Photograph by Tamara Reynolds

Teradata Magazine-September 2007