Credit Scorer Betting on Artificial Intelligence

Neuristics Corp., a new entrant in the credit scoring industry, claims its use of artificial intelligence technology will make its models stand out from the crowd.

The company boldly adds that its predictive systems will prove superior to those of its competitors - even Fair, Isaac & Co., which is by far the dominant player in the field.

"This stuff is 'Tom Clancy comes to credit cards,'" said James J. Carey, vice president of Northbrook, Ill.-based Neuristics.

Neuristics provides an example of how neural networks and other facets of artificial intelligence technology are finding their way into financial services marketing.

Banks are increasingly using neural networks for fraud control. The sophisticated software is well-suited to pattern recognition. It can discern from massive volumes of shopping and spending data minute variations that may indicate fraudulent activity, which can be brought to the attention of bank personnel.

Mr. Carey said Neuristics is the first company to use artificial intelligence in marketing models, which are designed to point card issuers in the direction of the most profitable customers.

He said Neuristics' models identify prospects who will not only respond favorably to a preapproved mail offer, but will use the card, revolve credit balances at the end of the month, and keep paying their bills on time.

"Everyone has finite resources," said Mr. Carey. Neuristics' array of products helps issuers "get the most for their marketing dollars by targeting the most profitable segments."

To build predictive models, Neuristics uses neural networks in combination with other forms of artificial intelligence, including fuzzy logic, genetic algorithms, and chaos theory.

The firm was established in 1993 by direct marketing expert Douglas McCrea and biomedical researcher Andrew Krause, who did pioneering artificial intelligence research at Johns Hopkins University. The staff includes 10 scientists who design models for clients.

Neuristics' first clients were retailers. Using predictive models, Neuristics guided two divisions of Pepsico - Taco Bell and Pizza Hut - to the most profitable sites for new locations.

The company said it has since snared several card-issuing customers, including First Union Corp., First USA Inc., Norwest Corp., and First of America Bank Corp. NationsBank Corp. said it is using Neuristics' services to build response models for its affinity banking group. Sears, Roebuck and Co. will use the models to cross-sell different products to its customers.

"What we find interesting and attractive about Neuristics' models is the ability to add the total profitability of the customer" to more traditional response and activation models, said Marc Altman, First of America's senior vice president of marketing.

Mr. Altman recently traveled to New York from his headquarters in Kalamazoo, Mich., to promote First of America's involvement with the company, though he said the bank has no financial interest in Neuristics.

Mr. Altman said the key to Neuristics' success is using all available techniques to create hybrid models. "When all you have is a hammer, all your problems look like nails," he said.

"Too often the tool defines the problem," said Neuristics' Mr. Carey. "We think the problem should define the problem."

Martin Sleath, vice president of corporate research and development at Fair, Isaac, agreed that "understanding the problem is key," he said. "Get the problem wrong and it doesn't matter what you hit it with."

Fair, Isaac employs various techniques in its models, including regression analysis, a proprietary approach to linear programming, and neural networks. But Mr. Sleath said the company hasn't found an application suitable for other types of artificial intelligence, such as fuzzy logic.

Jennifer L. Porter, spokeswoman for Credit Strategy Management, a credit scoring firm in Atlanta, said, "The power of the data dictates how powerful the model will be, not necessarily the technique used to model it."

She said many executives view neural networks as "the sexy new thing, but it has pitfalls like any other type of analysis."

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