Optifi Inc., a company formed last year by Fair, Isaac & Co. and the technology firm MarketSwitch, plans to bring a mortgage scoring product to market by yearend that can predict the likelihood that a loan will prepay before its stated maturity, much as Fair, Isaac's credit scores predict the probability of a borrower's defaulting on a loan.

Executives say the Edina, Minn., company's patented product, also to be called Optifi, will directly challenge the prepayment models that Wall Street, individual mortgage companies, and private technology companies use to price mortgage-backed securities and mortgage servicing for lenders and investors.

The five-person venture has some powerful backing. Its board includes Tom G. Grudnowski, president and chief executive officer of Fair, Isaac in San Rafael, Calif., and Drew Eginton, chairman, president, and CEO of MarketSwitch in Dulles, Va.

Donald E. Lange, Optifi's chairman and CEO, is a former president of the Mortgage Bankers Association who previously was president and CEO of Weyerhaeuser Mortgage and Weyerhaeuser Financial Services.

Optifi says its loan prepayment scoring product will differ from those used by competitors. Other products only analyze the loans' aggregate prepayment histories; Optifi's will look at loans on a loan-by-loan basis, the company said.

Furthermore, executives say their model builds on existing prepay models: In addition to using market information such as interest rates and portfolio loan data, the product will incorporate a wealth of demographic information, which should make its scores more predictive of how fast loans are prepaid.

The company says its product will incorporate hundreds of specific demographic elements for each loan, from the borrower's occupation to how much time the borrower has spent in the house and how many miles he or she drives to work each day.

Mr. Lange said the company initially plans to offer its product for pool valuations in servicing portfolio management. Ultimately, though, his vision is to use it to help calculate loan prices on a loan-by-loan basis, possibly at the point of sale, where it could help lenders price loan.

As Fair, Isaac does for its scoring products, Optifi will manage the data input and output to generate its scores.

Recent accounting standards have made it more difficult for lenders to determine the value of their servicing portfolios, and Optifi's product should help servicers come up with more accurate valuations, Mr. Lange said.

Officials also said the product will help lenders create more uniform pools for securitization and let buyers and sellers more accurately value pools of loans.

James P. Punishill, an analyst at Forrester Research in Cambridge, Mass., said that adding the demographic data is innovative. "Nobody else has put all of these databases and information together to give you one score."

But Michael Bykhovsky, president of Applied Financial Technology, a San Francisco company that sells prepayment prediction models for valuing securities and servicing rights, said that Optifi is not producing a prepayment model. "The only part of prepayment behavior they might be able to provide improved predictions of is the propensity for a borrower to move, but this is a less important component of prepayment risk. Your greatest risk comes from the propensity to refinance, which derives from interest rate changes and what type of loan the borrower is in. That information is well-covered by existing models."

Furthermore, he contended, a new player like Optifi, cannot hope to compete with established companies like his, which has had its model on the market for years. He said his product can also analyze loans on an individual basis.

But Wayne Atkins, Optifi's vice president for productization, said his company's model was generated using a lot of historical data and has already produced results. Optifi recently ran a study comparing its model with the internal prepay model of a "sophisticated, large" servicer, he said. "Ours was significantly better."

Ted Durant, manager of analytic services at Mortgage Guaranty Insurance Corp. of Milwaukee, said his company has been selling loan-level prepayment scores for four years.

MGIC's model was developed particularly for servicing and was formulated using loan-level mortgage servicing data, Mr. Durant said. Like other prepayment models, it relies on market and loan information but not on demographics, he said.

Mr. Durant, said using demographic data has a lot of potential for explaining prepayment behavior but that it is unclear whether anyone could ever realize this potential because of the legal issues involved. Only a limited amount of data can be obtained on any household, and it's expensive to acquire, he said.

Mr. Atkins said his company does not "intend for our score to be used for an acceptance or denial of credit" by banks. "We are very aware and cautious about issues surrounding the use of inappropriate variables like race, class, and gender, and we absolutely don't use those variables at all."

Optifi has licensed its product from MarketSwitch, which patented a model for prepayment propensity and duration probability. Mr. Lange said that Fair, Isaac and MarketSwitch contributed equally to the start-up, though he declined to say how much.

Meanwhile, Optifi is focusing primarily on how it will collect and distribute lenders' loan and pool information.

Mr. Lange said the company is also working on a marketing strategy and on identifying strategic partners and additional investors. He plans to approach his contacts in the mortgage industry, and Optifi is seeking to beta-test its product with lenders.

The company has a lead on a prospective testing partner and has been in discussions with several others, he said.

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.