Fleet Puts Muscle Into Building a Huge Warehouse

Fleet Financial Group is finding that building a data warehouse can require heavy lifting.

Consider the data emanating from the seven-state banking company's 2,050 automated teller machines. For three days earlier this year, Fleet technologists huddled to pore over the definitions of 450 data elements offered up by the ATM network. They went over every single one in deciding which to keep.

Among the 100 elements retained were those that showed the location of the ATM, the time of a transaction, the cash dispensed, and to whom. The other 350 categories were dropped because they were deemed merely "operational" and not suitable for inclusion in the vast data warehouse.

The 14-month, $38 million project has involved up to 60 full-time employees at the $83.4 billion-asset Fleet. At least five system integrators and technology companies-including KPMG Peat Marwick, Informix, Epsilon, Exchange Applications, and Sun Microsystems-are involved in the two-phase warehouse project. Fleet expects the system to be operational by Jan. 1.

Boston-based Fleet is hardly alone among major financial institutions seeking to build extensive-and expensive-vaults of customer and market data. (See accompanying box.)

Yet Fleet's effort offers insights into how a bank that has already grown broader through mergers and acquisitions may be able to take further advantage of its scale through data warehousing.

How Fleet conceived, organized, and plans to implement its data resource can offer clues for other banks trying to figure out what works and what doesn't in building a warehouse.

As recently as a decade ago, Fleet was just a Rhode Island bank with $12 billion of assets, 44 branches, and 51 ATMs. Through acquisitions of Norstar Bancorp, Bank of New England, and most recently Shawmut National Corp. and National Westminster Bancorp, Fleet grew into the 11th-largest bank holding company, with more than 1,200 branches.

With the acquisitions digested, "we are not interested in market gain," said Robert B. Hedges, senior vice president of the direct financial services group at Boston-based Fleet. "Instead, we want to deepen relations with our customers."

"Having played out the acquisitions and consolidations, the imperative is to build on the franchise," said Randall B. Grossman, senior vice president and director of customer data management and analysis.

Fleet officials see data warehousing and data mining as the tools they need to reach further into the wallets of 10 million customers in six million households.

"The area I manage was created 18 months ago to build a world-class capability in data base marketing, customer behavior analysis, and ultimately delivery of the information" in a form that can help the bank expand its customer relationships, said Mr. Grossman.

Mr. Grossman said the thinking about data warehousing has changed from the 1980s, when the quest was for "a massive, centralized system that everyone in the bank could draw on."

"That concept has largely been set aside," he said, "because it was responsible for a lot of large and not-so-successful projects."

The real value of data warehousing in the financial industry, most experts agree, is its capacity for analyzing customer behavior and tailoring marketing strategies appropriately.

In the beginning, many bankers believed they had to build huge data repositories first, then figure out how to apply them. That pendulum has swung to the point where banks are starting small-in Fleet's view, perhaps too small.

"The conventional wisdom is that the safest way to do this is to build incrementally," said Mr. Grossman. By not getting caught up in "meta- architecture" and beginning with a simpler and more specific "data mart," a bank can show positive returns within a year, this theory holds.

The advantage in this approach is that even a single division of a bank can derive benefits, regardless of whether top management and overall strategy embraces the concept for the entire bank.

The downside, which Fleet sought to avoid, was that discrete projects prevent banks from realizing "customer knowledge" across all departments and product lines.

"We have quite consciously chosen to make a much larger up-front investment in an effort to build the infrastructure to get far more benefits earlier in the process," said Mr. Grossman. "The problem with projects the size of Fleet's data warehouse is that if you fail, firing is not enough revenge."

Fleet technologists necessarily had to get top-management to buy in at the beginning. A relatively rigorous examination of the bank's business needs followed, Mr. Grossman said.

Just as Fleet's ATM systems yielded a wealth of data, some of which was warehoused while the rest was left in mainframe computers, information from some 65 other transaction systems needed to be pruned. These systems, the guts of a financial institution's technology, are used to track teller activity, deposits, loans, and other processes. Some 2,000 activities were planned over a 14-month gestation period.

While all of these computer systems were engineered for efficiency, Fleet found that warehousing was essential to dig deeper in understanding and potentially even predicting customers' behavior.

Fleet has a plan for two data marts to take advantage of the warehouse. Phase 1 would provide customer analysis for use primarily in marketing campaigns. The second phase would produce other management information.

The marketing information can, for example, help Fleet prune from promotional mailings for its toll-free telephone service the people not deemed likely to use telephone banking.

By aggregating every single ATM transaction, the bank can begin to understand which neighborhoods are most used by their customers and which machines owned by other banks are popular with Fleet customers.

And more accurate tracking of which sorts of customers frequent which branches can help Fleet better focus its mutual fund sales force.

Originally scheduled for completion by the end of October, Fleet's first phase one is running three to four weeks behind schedule, Mr. Grossman said.

After a 45-day pilot, Fleet will begin with 200 users. It expects to have 2,000 within two years.

That second phase will lead to a third data mart sometime in 1998 that will deliver selected information to branch staff and telephone representatives. The new data mart will permit more refined ratings of potential customer profitability.

"Because we are beginning to collect a massive amount of information, we have to be very conscious about how to ensure the privacy of information we collect," said Mr. Grossman. "We have made it clear that looking at information that you are not supposed to is grounds for dismissal."

All told, Fleet's warehouse will hold a million megabytes-one terabyte- of data, enough memory to track all information in customer accounts for 36 months. Yet the scope of Fleet's warehouse is small in comparison to Citicorp's seven terabytes and Sears Credit's 2.5 terabytes.

But Fleet expects to build additional applications based on the knowledge it gains as it begins to put its warehouse to use.

"In truth, we will never be done with it," said Mr. Grossman.

For reprint and licensing requests for this article, click here.
MORE FROM AMERICAN BANKER