At an afternoon meeting in August of 1994, Jack Sweeney, Bank of Boston's head of information management resources, gave vice chairmen Edward A. O'Neill and William Shea some troubling news.

Mr. Sweeney told the vice chairmen unless all the managers in the bank got together and agreed on some basic ground rules, the million-dollar systems project that had been underway for a year and a half would not work.

The project, to build a data warehouse that would store in a single place information from many different areas of the bank, was intended to give the bank's decision makers the means to better analyze trends and determine risk and profitability.

The goal of building a data warehouse is straightforward: extract data from legacy application systems across the bank, and put that data into a single relational database.

The arrangement lets line workers continue to use the data in the legacy systems, while giving decision makers - the risk managers, the marketers, the corporate finance departments - easy, centralized access to corporate data.

But, Mr. Sweeney argued that putting all the data in one place was not going to be enough, because the data from one unit of the $45 billion asset bank could not be understood by a banker from another unit.

To illustrate his point, he displayed a list of accounts for a single corporate customer. The customer had 13 accounts with Bank of Boston, yet the customer was identified by ten different names, 13 different customer numbers, and eight different addresses.

If the project went on as before, "you'd have a data warehouse for finance, one for small business, and one for cash management," said Mr. Sweeney, recalling the meeting. "You'd just have additional islands of automation."

The meeting had a tangible outcome.

"We realized that Jack had a good firm handle on where he was going, and that the business units needed to take hold of the opportunity to use this new tool," said Edward Dolan, director of financial projects at the bank, and one of the meeting's attendees.

The group decided to set up what it calls a Corporate Data Resource Board, a group of 30 business line managers who get together once every three weeks to hammer out a consensus on such deceptively simple topics as: What is a customer? What are the bank's products? How is the bank organized? Who is responsible for the data?

Mr. Sweeney chairs the meetings. For the business unit managers, unfamiliar to systems projects, the process can be complicated. "Actually creating the tables of information is more difficult than I'd have anticipated, because of the sheer volume of data," Mr. Dolan said.

Once consensus is reached, about a dozen technologists work on extracting the right data from legacy systems, and putting it into the warehouse.

Initiatives such as this have made Bank of Boston "one of the leaders" in data warehousing, said Rick Makos, director of worldwide marketing for financial services decision support products at Unisys Corp., Blue Bell,Pa. "Mr. Sweeney has a very viable strategy."

Mr. Makos likened the bank's approach to its warehouse to a library with a card catalog. Instead of going through book after book, you let the index file direct you to the right book.

Although data warehousing is now a hot topic, a few institutions, like Fidelity Investments and BankAmerica Corp., have had warehouses for many years.

Studies by the Advisory Board, a financial services research group based in Washington, found that credit card companies that use corporate decision support systems - usually supported by a data warehouse - have double the return on assets of the average credit card company.

The uses of a data warehouse are varied. On Wall Street, investment banks want to be able to identify their risks and exposures at a given time, but the data is locked up in five or six different data bases in different locations. Special data extraction software obtains data from the databases and brings it all together in a data warehouse, where it is reformatted.

Commercial banks can use a data warehouse to quickly calculate product and customer profitability. In derivatives trading, for example, a decision-support system based on a data warehouse can take into account hundreds of variables in analyzing risk factors.

One of the most compelling reasons for building a data warehouse is to help a bank do sophisticated target marketing. Not only does the warehouse collect all the data the bank already has about its customers, but each marketing campaign that is done adds more information to the profiles built up in the warehouse.

The technique can be continually refined, based on the knowledge gained from a campaign, Mr. Makos of Unisys said.

The projects themselves range in cost from under $500,000 to over $6 million, depending on the size of the computer used. And they require a commitment of many years.

At Bank of Boston, the initial driver of the project was Michael Simmons, chief technology officer at Bank of Boston in the early 1990s, who had successfully implemented Bank of America's data warehouse in the 1980s. Mr. Simmons left Bank of Boston in 1993, but top management remained committed to the concept of a data warehouse.

Vice chairmen O'Neill and Shea gave strong backing, Mr. Sweeney said.

"In building the architecture (for the warehouse) we were putting time and money and resources into something that doesn't have an instant deliverable," Mr. Sweeney said. "We've got leadership that has vision."

The project has now been underway for over two years. Currently the warehouse resides on an IBM mainframe on the DB2 relational database, but Bank of Boston has requested this year for a system called a symmetric or massively parallel processor - a computer system, similar to a supercomputer, that can handle huge volumes of complex queries more quickly than can a mainframe.

"Right now we don't need a massively parallel processor, but we know that by next year, the DB2 on the mainframe won't cut it."

Seven staffers work full time on the data warehouse infrastructure. Another five employees are devoted to extracting data for a small business lending unit project, the bank's first test of the warehouse for a business unit. In the small business area, the warehouse initially will be used for target marketing and for trend analysis. Marketing managers will be able to look at existing households, see what kind of products these customers buy now, then use those households as a model for what similar households might buy.

The pilot also will be used to look at the growth of customers and households, the degree of customer retention, and the incidence of product cross-sells.

The bank said it expects to have the pilot up and running by the second quarter of next year.

"There's no such thing as a small pilot of a warehouse," said Danielle Barr, director of MIS for customer strategies and insights. "You can't just build something for small business without rationalizing the products for the whole bank."

The group eventually decided on several levels of customers: the individual account, the customer or corporation, the household or parent company, and groups of households or super-parent companies.

"The important thing is that the warehouse can function for all groups, without having to build a new structure" for each unit of the bank, Ms. Barr said.

Mr. Sweeney contends that the most important thing learned in the process of building a data warehouse was a lesson from last August.

"It's not enough to have a warehouse," he says. "The most important thing is the quality of the data you put in - not just data integrity, but data consistency." v

Ms. Brokaw is a freelance writer based in San Francisco

Subscribe Now

Access to authoritative analysis and perspective and our data-driven report series.

14-Day Free Trial

No credit card required. Complete access to articles, breaking news and industry data.