Fleet Financial Group Inc., Providence, R.I., going companywide with neural network software intended to help it tailor its cross-selling efforts to the needs of individual customers.

Neural networks are computer systems designed to mimic the problem- solving process of the human brain.

Bank officials say Fleet's use of the technology is part of a shift to a more aggressive sales culture in which bank employees try to anticipate customer needs.

"Despite the fact that a lot of people talk about taking a customer's viewpoint, very few actually do that," said John Menke, a vice president with the $48 billion-asset bank.

During the six-month-long test of the system, Fleet worked with Harte Hanks Data Technologies, based in Billerica, Mass., and Neuralware Inc. of Pittsburgh.

Mr. Menke said one of the benefits of the technology is that it centralizes cross-selling data, thus helping to eliminate counter- productive marketing efforts waged by separate product groups.

At many institutions - Fleet included - different product groups often confuse customers with promotional materials that give conflicting messages.

For example, one division of the bank may launch an advertisement that says, "Save now; it's a great time to get high CD rates." The same bank may also run a second promotional campaign that says, "Traditional bank investments are no good, open up a Galaxy fund," Mr. Menke said.

"If you don't have something that acts as the score buster, you are going to send conflicting messages," he said.

Fleet officials said the new sales strategy will help it compete against such nonbank financial service providers as Fidelity Investments and Merrill Lynch & Co.

"It's not a game of gaining new households; it's a game of hanging on to those that are currently profitable," Mr. Menke said.

Aaron Knott, director of special projects with Harte Hanks, said neural networks are a potent tool for cross-selling bank products because they enable banks to "tailor information to their clients based on what the client is interested in."

Experts said the banking industry's use of neural networks historically has been concentrated on fraud prevention in the credit card arena. The technology also has helped predict the likelihood of delinquencies, the need for increased credit lines, and whether a customer will want additional credit cards.

Sushmito Ghosh, a vice president with Nestor Inc., a Providence, R.I.- based software developer and consulting firm, said neural networks are also used in character recognition technology for signature, check, and other types of paper processing.

Patricia McGinnis, an analyst with Tower Group, Welleseley, Mass., said Fleet's use of neural network software is appropriate because the technology "is best for complex pattern matching environments, and the many characteristics of an individual consumer's financial and business interests do in fact constitute complex patterns."

Neural networks have been around for a dozen years, but after an initial flurry of interest, many financial institutions backed off the technology as they encountered difficulties in applying it to their operations, Ms. McGinnis said.

Still several banks use the networks in the small business markets and for other "experimental systems being developed for various purposes," she said.

"It's receiving wider acceptance now than in the past because the technology has proved itself in real-world applications," added Mr. Ghosh. "It's on the threshold of widespread adoption."

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.