How data mining leads to loans for the underbanked

While Big Data is increasingly a passion of forward-thinking companies looking to extract value from a growing pile of information, it is also having an effect on some of the most vulnerable consumers, who are finding a way to secure loans thanks to data mining. That’s through the work of ZestCash, a next generation loan service, that has been applying data creatively to bring loans to consumers who have few choices beyond payday loans.

Douglas Merrill of ZestCast at Structure:Data 2012

(c) 2012 Pinar Ozger. [email protected]

ZestCash CEO and founder Douglas Merrill, who formerly served as CIO at Google (s goog) appeared at GigaOM’s Structure:Data conference to talk about how ZestCash uses a wide array of data to help qualify people for short term loans. The company, which launched in 2010 and raised $19 million last year, has been giving $300 to $800 loans to users based on thousands of variables, which are boiled down to 10 models. Merrill is cagey about which data points are used but he said ZestCash does not factor in FICO, the credit score most often applied by lenders.
FICO uses 15 factors to determine who gets a loan, but that often leaves out at least 15 percent of consumers, who don’t qualify according to those standards. But by drawing in a lot of online data, ZestCash is better able to predict if a consumer will pay back their loan. The company now boasts a default rate well below half of the 40 percent rate in current loans.
For example, Merrill said by examining a user’s pre-paid phone account and checking for defaults, they can get a much better sense of whether they are likely to default on a loan. “If they lose (their phone), they lose their connection to the world. It’s a hugely important credit signal,” Merrill said.
By being smarter about which data it uses to rate consumers, ZestCash is able to charge half the fees of traditional payday loans. That can mean $600 more a month for a consumer, he said.
For now, Merrill thinks the ZestCash model will focus on underbanked users, who can’t get access to credit cards. Ultimately, he thinks ZestCash’s techniques — which he calls “big inference”, not big data — can be applied to all credit cards users.
“I think that approach will be so powerful over the next 5-10 years, you’ll see everyone migrating to this big inference approach,” Merrill said.
Watch the livestream of Structure:Data here.