Innovation Speaker Douglas Merrill On Credit Risk: Big Data > Traditional Math
Banks and lenders need to take into account a person's financial background to assess the likelihood of being paid back, and the risks involved with doing business with someone based on previous credit history. However, there is a human element missing to the whole equation, Merrill says. Sometimes bankruptcy, for example, matters in terms of whether someone will be able to repay their next loan, and sometimes it doesn't. Traditional assessments, however, don't dig that deep.
Merrill just unveiled the newest upgrade to his underwriting software at ZestFinace, which tackles this problem. Hilbert, as it's called, uses ten separate models paired with 70,000 signals to assess hundreds of variables surrounding a borrower's credit history. "The 10 models vote in a way," he explains, "it's like getting your 10 smartest friends around a table and asking their opinion about something." So far, the new model shows a 50% improvement over his last model. And the previous model (Hollerith) was about 50% better at assessing approval rate and default than the industry average. By using the innovative new math techniques he learned at Google (where he was the Chief Information Officer) and fine-tuning them, Merrill is able to provide more accurate credit assessments for those in need. "The ultimate win will be when major banks realize this is a huge win for them," Merrill says. It is innovative, forward-thinking like this that has made Merrill so successful. In his talks, he teaches companies how to harness new technologies and cultivate cultures of innovative to advance—and win—in their industry.