Credit Scoring: Models: Use 'Em or Lose It

Given rising consumer defaults, it looks as if a good many banks, particularly credit card issuers, are tossing caution-and predictive modeling-to the wind.

Not for long, though. Market conditions will likely snap bankers back into reality, reminding them that predictive modeling creates consistent, predictable portfolios, which translate into lower delinquencies and defaults, and, ultimately, a better spread over funds.

One way to create this lenders' Nirvana is by using a predictive model. Micheline Martin, vp of credit scoring at Toronto's Royal Bank of Canada is a big believer in models that are purchased instead of those that are developed internally at the bank. "We think that training the people internally (to create and maintain predictive models) would be more costly than to outsource (it)," she says. "And then we become vulnerable because they could work for me one day and then go tomorrow and work for a major competitor. So I'd prefer having an independent supplier that I can rely on."

Martin says that, while she can't quantify how much she saves the bank by outsourcing to Atlanta-based CCN Inc., the real pay off is in performance consistency. "I've never made a business case for it," she says. "And even if I (did) find out that I could save a few bucks by doing it internally, I'd still outsource. There's no price you can put on losing an analyst. How much (money) are you losing every time you lose someone and have to train someone else, and seeing your strategic information going from your place to your competitor's?"

Martin says the main benefit from using the CCN model is creating consistency in the bank's collections and in reducing delinquencies. But she says that Royal Bank is finding other uses for it, too."We already use it for marketing and cross-selling."

-reinbach tfn.com