Editor's note: This is adapted from a post that originally appeared on LinkedIn.

The financial press has offered up stories recently of new lending startups relying on alternatives to Fair Isaac in scoring credit applicants.

This type of evolution in consumer finance is inevitable. A tool like Fair Isaac’s FICO scores is the dominant measure of a borrower’s credit eligibility. But it is healthy to see alternative lenders seek new types of data, and use more processing and analytical capability, to improve loss prediction – particularly for deserving “thin-file” borrowers who are ill-served by Fair Isaac. After all, Bill Fair and Earl Isaac did the same disruptive thing to the once-dominant rule-of-thumb consumer underwriting practices years ago.

But shareholder and loan buyer beware. There's not much history to assess the performance of revised scoring inputs favored by marketplace and other alternative lenders. We won’t really know the predictive value of the new measures until well into the next recession. Wall Street's continuing to require lenders to provide parallel FICO scoring is therefore prudent, at least until the new predictive metrics have been truly tested in a downturn.

Formal consumer credit scoring began 60 years ago when Fair Isaac’s two founders each invested $400 in the venture. They used a borrowed computer that was probably the entire size of the Northern California apartment from which they ran the company.

FICO was the first consumer credit scoring tool, and it still represents the gold standard, despite some industry misunderstanding about what credit scores actually predict and a lingering tendency on Wall Street to underestimate how quickly they can change (as we learned during the last financial crisis). It’s still the most effective mass-market credit-scoring tool, although customized scoring approaches can work better for small niche lending products and markets.

I helped take Fair Isaac public as an investment bank lawyer in 1986. Then CEO Bill Fair was the very model of an operations research guru from Stanford Research Institute. He looked like a NASA engineer, with the messianic intelligence, white short-sleeved shirt, trim haircut and pocket protector to prove it. He knew the math, and the math told him that he was on the leading edge of a revolution in predictive analytics (in many ways the first use of "big data" in banking).

In the early 2000s, I helped run banks that relied heavily on FICO for credit decisions as they made consumer loans available to the mass market. So I know firsthand how central FICO credit scoring has been to the consumer lending revolution, for better and sometimes for worse.

But new lending startups have not been shy about wanting to upend the FICO model. Leading online lender Social Finance Inc. made no bones about the company’s opinion of the traditional scoring method when it branded itself the first “FICO-Free Zone” in a press release last month. “Banishing traditional credit scores stems from SoFi’s belief that the FICO model is flawed and outdated,” SoFi’s press release said.

The release quoted CEO and co-founder Mike Cagney as saying, “We found that the FICO score was anything but transparent. So we threw it out.” Instead, SoFi said that it will consider three criteria — employment history, track record of meeting financial obligations and monthly cash flow minus expenses – to determine if an applicant is qualified for its consumer loan products, which are marketed to ambitious and successful recent college and grad school alumni.

Others, like Avant, an online lender on the high-risk side of the credit spectrum, say they never relied on FICO anyway. Avant uses FICO scores only to screen out those applicants with scores below 560 (deep in subprime land) and then applies its proprietary models to make credit decisions.

In the abstract, no one can argue with using things like employment, cash flow and payment history as predictors of credit quality. But the devil is in the details when building effective scoring models. Credit-score-building requires an equal number of "good" and "bad" borrowers for validation and, as Ann Rutledge of R&R Consulting has pointed out, the mathematics of scoring involves special mapping techniques that you will not know unless you have been taught formally by someone with deep credit-scoring experience.

Originating $6 billion of loans – $5 billion in 2015 alone – that have performed well during an economic upswing, by itself, shouldn’t give SoFi confidence in its new scoring system. As much as we would like the opposite to be true, we need to acknowledge that credit scoring is an inexact science which uses hard data in a quest to predict what we soft, illogical, biological humans will do in a financial crunch.

Todd H. Baker has been chief strategy and development officer at three of the largest U.S. banks and a partner at two leading international law firms. He is currently the managing principal at Broadmoor Consulting LLC.