Visa, Dun & Bradstreet Launch New Modeling Tools

Financial services provider Visa U.S.A., San Francisco, and database concern Dun & Bradstreet, Murray Hill, NJ, yesterday unveiled two new predictive modeling tools designed to help Visa business credit card issuers market more effectively.

The new tools, the Visa Business Prescreen Model and the Visa Business Underwriting Scorecard, can be used in combination with the Visa Business Response Model, a tool developed by Visa and Dun & Bradstreet in 1997 that assists banks in identifying prospects for business credit card offerings.

The Prescreen Model, which is available immediately, is designed to identify small business prospects that have an acceptable risk profile before sending them an offer. It predicts which businesses are likely to have a severe credit card delinquency during the next 12 months. The Underwriting Scorecard, which will be available early next year, is used to evaluate applicant risk during the underwriting process.

The models leverage Dun & Bradstreet’s credit information database of 11 million active U.S. business records, combined with actual business card performance data provided by a group of Visa business card issuers. The Prescreen Model, which was built using 25 variables obtained from accounts that are at least 18 months old, was compiled from nine contributing card issuers that were representative of small businesses throughout the country, according to Ann Brisk, director of Visa Business.

“One of the reasons this is such a powerful tool is that it’s the first screen based entirely on small business credit cards,” said Brisk. “That’s pretty unique in that most of the others have been based on other trade references outside the credit card environment.”

Visa charges issuers for the service based on a price per thousand names that they run through the model.

Brisk said she expects that out of about 200 total Visa card issuers, 20-30 will use the new models. Currently seven issuers are testing the Prescreen Model by running past mailing lists through it to see how well it would have predicted delinquency.

“Most of our issuers have not been in the market long enough nor do they have large enough portfolios to build really accurate custom models,” she said. “So, essentially, by using pooled data, they are able to take advantage of an aggregate of data that’s representative of a nationwide sample, vs. just a regional sample. It’s especially useful for an issuer that’s going to be issuing outside their regional footprint.”

Related Posts