As data warehouses gain in popularity, financial institutions are typically taking one of two approaches to installing them.

The first involves developing a framework for how data will be stored and integrated across the organization before planning how individual bank divisions will use the information.

"Ultimately, it would be nice to sit back and architect an enterprise data warehouse, an enterprise data model from day one," said Richard Rist, vice president of research at the Gaithersburg, Md.-based Data Warehousing Institute.

But with this approach, "you would literally spend a couple of years in development," he said.

The alternative path, which most banks seem to be taking, is working on the enterprisewide architecture of the warehouse while constructing repositories for different divisions.

But this approach also has some drawbacks, among them the possibility that the repositories will not tie together as expected, making the enterprisewide warehouse less comprehensive.

At least one bank, BankBoston Corp., is trying to meld both approaches.

Its effort is springing, in part, from the acquisition of BayBanks Inc.

BankBoston is "leveraging some of the enterprisewide architectural efforts at Bank of Boston with the more targeted data warehouse or data- mart development we undertook at BayBanks," said Avi Gusman, director of corporate data management.

The effort began in 1994, when BankBoston started talking about how to make its decentralized data available to large portions of its work force.

To start building the warehouse BankBoston turned to Burlington, Mass.- based Cayenne Software Inc.

The bank had been a customer of Cayenne's predecessor Bachman Information Systems. (Cayenne was created last July through a merger of Bachman Information, a development tools company, and Cadre Technologies, a data analysis and design specialist.)

"We're using Cayenne's technology to actually build the data model for the warehouse," said Peter L. Smith, senior section manager with the bank's corporate data management.

"We're approaching it from an enterprise perspective so that we have a single corporate standard for key pieces of business information. The overall objectives are to standardize and to increase the efficiency and accuracy of management information reporting," he added.

But the bank sees other potential applications for the harvested data.

"For retail banking, we have a treasure trove of information in our transactions," said Mr. Gusman.

"We intend to leverage the infrastructure more actively with data mining technologies to enhance our segmentation, profitability analysis, and- potentially-cross-selling," Mr. Gusman added.

One of the first steps in constructing an enterprisewide data warehouse is creating naming standards. For a warehouse to work effectively, terms like "customer" must mean the same thing bankwide.

Developing naming standards is part of creating what is often referred to as the meta-data-a kind of index that helps users navigate a warehouse.

Also crucial to the early stages of construction is sharing information between departments. "Every division wants information from another, but very few are willing to unlock and allow other areas to get access to theirs," Mr. Rist said.

Other hurdles are handling end-user expectations, and defining and managing a project's scope, Mr. Gusman said.

"Expectation management is really the critical factor in getting some perception of success," Mr. Gusman said.

BankBoston expects to see the fruits of its labors soon. The central data warehouse is expected to be completed by midyear, Mr. Gusman said.

Other institutions eager for clues about how to implement a data warehouse projects are expected to watch projects like BankBoston's closely.

Having a detailed plan in place before making a financial commitment is crucial, said Paul Fluckiger, president and chief executive officer of Carleton Corp., a Billerica, Mass.-based vendor of data warehouse technology.

Mistakes could be costly. For banks with assets between $1 billion and $4 billion, an enterprisewide data warehouse costs around $3 million to plan, design and implement, according to Mentis Corp.

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.