Upstart co-founder Paul Gu

Upstart ready to transform underwriting

Upstart co-founder Paul Gu
Upstart co-founder Paul Gu
Underwriting methods have a long way to go before they are accurate predictors of human behavior,Upstart’s Paul Gu believes.

Mr. Gu is a featured speaker at Crowdnetic’s Crowdfinance 2015 later this week in New York City.

He will be addressing the topic of borrower risk.

“At Upstart, we think about how to make underwriting much better than it is today,” Mr. Gu said.

There has been little innovation in consumer credit, save for some small improvements in FICO, Mr. Gu said. The traditional approach is to analyse the track record of someone making long-term payments and extrapolate from there.

Gu says,”While it’s effective, it is far from perfect.”

He’s referring to the upside-down bell curve that puts most borrowers on either extreme. But that doesn’t reflect reality.

“Most credit scores are in the middle,” Mr. Gu explained. “That means we don’t know if they will pay back or not.”

“There is this natural tendency to think (the normal bell curve) is normal distribution,” Mr. Gu added. “But what you want is bipolar distribution where each peak is at opposite ends of the spectrum.

This means traditional scoring methods are insufficient, and the universe of variables considered in the underwriting process have to expand.

Mr. Gu said the scoring issue can affect people early in their adult lives – a problem frequently experienced by millennials.

“Graduates do not have a credit history needed to engage in many things. Getting an auto loan and even renting can be difficult.”

This problem disproportionately affects people in the middle- and lower -ncome levels, Mr. Gu explained. Whereas, in higher-income families children are placed on a parent’s credit card at a young age, so when they are 22 and graduating from college they may already have a decade-long credit history.

Compared to other parts of the world, Mr. Gu agrees we have it good on the credit front. Credit-reporting systems are fairly robust and agencies play an important role.

“The key takeaway is the credit-scoring system is not that good,” he said.

“It is easy to find bad credit scores, But the ability to identify good risks with a good explanation of why, is low. There are plenty of false negatives and some false positives.”

He knows one-time events are given disproportionate weight by traditional scoring systems, even though those events have low predictive value.

That’s why Upstart finds the field so challenging and ripe for change. Something Mr. Gu will be
sharing at the conference in New York City.