As Big Data continues to evolve and become part of many companies’ everyday tools, one area that is seeing significant impact is the lending industry. There are thousands of data points lenders can use in order to determine the creditworthiness of an enterprise.
Over the past several years, tech-focused firms have been mining data in order to give a more complete picture of young businesses’ financial health, and that is helping to bring credit to business owners who have limited or no credit histories. We caught up with Douglas Merrill, the founder and CEO of ZestFinance, based in California, to get a sense where Big Data underwriting may be headed.
You take a very interesting approach to underwriting by using machine learning and large-scale big data analysis. How does your technology work, and what has inspired this approach?
DM: We analyze tens of thousands of variables created by machine-learning algorithms, modify them based on patterns, trends and unique insights and feed the modified variables into multiple Big Data models.
This is in contrast to the 10–15 variables that “traditional” scoring models use. When you analyze trends across hundreds or thousands of signals, they don’t create a lot of value individually, but if you get enough of them and string them together, you get very powerful information.
What are the tangible benefits that business lenders can achieve by using your model?
DM: Our technology enables lenders to better assess credit risk, expand credit availability and achieve higher repayment rates.
When ZestFinance first entered the market, your goal was to save the underbanked millions of dollars. How successful have you been in this regard, and how have your goals evolved over time?
DM: Our mission is to provide fair and transparent credit around the world — with a particular focus on people who are underserved. With the recent announcement of our joint venture with JD.com in China, we will provide credit evaluation services to lenders so that they can extend credit to Chinese consumers who don’t have credit histories and therefore currently don’t have access to credit.
We will continue to apply our technology to transform consumer credit in markets that make sense and where we can have the biggest impact.
What are the top misconceptions in the market about Big Data, and how would you address it?
DM: Many people and companies think success in data science is all about Big Data — being able to manipulate extremely large volumes of information and scale black-box algorithms. This is not always the case. Getting data science right isn’t a matter of the volume of data you have. It’s about understanding the quality of data you’re working with and knowing exactly what to look for in that data set.
Here’s a quote by you: “We need data artists to save Zombie borrowers.” Would you kindly elaborate a bit more?
DM: Data artists are individuals who can interpret the intricacies and complexities of data and build complex models to predict human behavior. This becomes incredibly useful when analyzing the data of “zombie borrowers” — those who show up as “dead” in their credit on their credit bureau files but in actuality are still alive and the data is wrong in their credit bureau record.
Using a combination of machine-learning algorithms and data artistry, we are able to identify errors like this and create a full picture of a person’s creditworthiness despite missing and inaccurate data. This enables us to provide fair and transparent credit to people — and help other lenders do the same — because when you are able to effectively predict creditworthiness, you can drive the cost of credit down.