September 27, 2004
(Computerworld)
When the stock market slumped in mid-2000, most financial services firms found themselves struggling to hold onto clients. ING was focused on selling them new products.
"Not only must large companies retain customers, but to increase revenues, they have to cross-sell to them more effectively," says Linda Marr, chief knowledge officer at Atlanta-based ING Americas, a division of Amsterdam-based ING Groep NV.
To help it draw upon customer data across multiple accounts more effectively, ING embarked on a massive data integration effort in July 2001 aimed at creating a series of database "hubs" that together form a unified information architecture for its seven U.S. business units, including reinsurance, pension plans and mutual funds.
"The hubs are for centrally structured data ... enterprise value data that's commonly used across all lines of business, [including] customer information, agreement information and product information," says Chief Technology Officer Raymond Karrenbauer.
ING, which has 27 lines of business worldwide, has completed four hubs and has 10 more under construction. Those multiterabyte databases give business units a comprehensive, unified view of the customer. But Karrenbauer had two broader goals in mind: create a reusable architecture to drive costs down and mitigate risk.
To appreciate the immensity of this project, one needs only to walk into ING's project control room in Atlanta. Tacked along the walls are dozens of schematic diagrams that illustrate how the financial services firm is simplifying an incredibly complex IT environment. That system was created by a flurry of mergers and acquisitions that included the purchase of Barings PLC in 1995 and Aetna Financial Services Inc. in 2000.
Raymond Karrenbauer, CTO at ING
ING also uses data profiling software from San Francisco-based Evoke Software Corp. and ad hoc reporting and analytical tools from France-based Business Objects SA.
In just 11 months, ING's staff built an object-based database infrastructure that melds unstructured and structured data from 2,000 databases into one logical view of customer and operational information.
Now, when an ING call center agent pulls up a customer record in the new database, "it will retrieve the customer information from the structured data and then also provide any relative documents with that customer," says Karrenbauer.
ING created DB2 repositories to aggregate a subset of structured data from each business unit's database that's of interest to the other business units. Along the way, the source data is funneled through a hub that conducts a series of transformation and deduplication steps.
Unstructured data sits in a separate repository. Sitting above that is an architectural layer that uses metadata to determine common traits between structured and unstructured data, says Karrenbauer. Some data is transferred in batch mode, but ING is piloting real-time, bidirectional data updates between the databases.
The source databases used by ING—which include Ingres, Sybase, SQL Server, Oracle and DB2—remain in use for claims and statement processing, says Karrenbauer. Under the data integration effort, ING merely centralized some of the source data from these systems and classified it using common terminology so information can be shared across business units.
Avoiding Duplication
Karrenbauer places the hardware costs associated with each information hub at $2 million to $4 million, but he says some of the projects have yielded an internal rate of return in excess of 100% by avoiding project delivery duplication. That's because ING created a standardized project delivery model that can be repeated as each business unit pursues its own relational databases, according to Karrenbauer.
Although reusability leads to lower costs, "there's risk mitigation in taking this approach," he says. "If you did one-offs in each case, the risks are clearly higher."
And as other ING divisions in Brazil, Chile, Canada and the Netherlands apply the information-hub approach adopted by ING Americas as a template for their own data integration efforts, Karrenbauer estimates that they can reuse 50% to 65% of the IT architecture that his division created.
ING's efforts have won praise from industry observers. "I've seen other companies try to stitch together pieces of applications into new applications, but ING is a real leader in its formation of an integration competency center and its use of data services," says Philip Russom, an analyst at Cambridge, Mass.-based Forrester Research Inc.
Marr says the unified IT architecture is already paying off. For instance, she points to recently enacted antispam regulation that requires financial services companies to offer customers an e-mail opt-out option. By now having all of its customer information accessible from a single environment, she says, ING is about to comply with the regulation "while allowing us to continue to market to and support our customers."
ING’s Integrated Data Environment
ING collects customer data from an array of 2,000 databases scattered across 27 business units. Information from source databases is cleaned, transformed and deduplicated in the information hub before passing to the consolidated DB2 UDB EEE databases running on a pSeries 690 multiprocessor server. The new database gives ING’s business divisions the ability to reuse customer data more effectively. Unstructured and semistructured application data is routed to the IBM Content Manager server where it is accessible based on metadata tags.
