The online lending industry is going through rapid changes with the advent of big data and machine learning in processing loan applications. Krishna discusses OnDeck’s workings.
Sramana Mitra: Let’s start with talking a bit about yourself in setting some context, and also introduce our audience to what OnDeck is doing in the realm of Big Data.
Krishna Venkatraman: Let me tell you a bit about myself. I joined OnDeck in October of 2013 as the leader for the data analytics organisation. I hadn’t really heard of OnDeck until I first talked to them. I found their value proposition fascinating. There was this huge unmet need among small businesses when they were looking for capital.
When I talked to the company, I realized how much of that problem could be remediated through the intelligent use of data, which is what attracted me to this role. I’ve been with the company since then. That’s almost three years. In that time, we’ve built up a formidable data organization within the company of over 65 data scientists, data engineers, and a lot of specialized expertise around different disciplines that matter.
We focus on technical depth because we really think that the specific disciplines that we are interested in have to be really strong and fundamental to what we build. You have to accompany that with a lot of domain knowledge and intelligence around how to apply data intelligence to improve decisions. We try to combine those. The founders, a long time ago, had identified this need.
If you think about small businesses, over 40% of businesses get nothing at all or get far less than what they wanted when they source funding. That’s a persistent problem that’s only getting worse. It seems to be endemic. It seems to be persistent. The question is why that is happening. We think it’s because of the difficulty of identifying the need that businesses have. There’s about 5.4 to 5.7 million small businesses that actually have employees and about 23 million sole proprietors. It’s very hard to get information about that on a consistent basis. Unless you’ve built an information engine that can actually understand this businesses very easily, it’s hard to actually serve them well. The company has taken on that challenge of trying to build products, services, and technology in support of meeting that need.
Sramana Mitra: Talk a bit about how you do that. This is a very large number of small businesses that you need to understand. You need to understand their credit-worthiness. You need to reasonably predict their ability to repay these loans. What data do you look at? How do you determine what you determine? Do some use cases and help us get a visceral feel for how you do what you do.
Krishna Venkatraman: It actually starts with having the right information in place and having that data organized, and then be able to make decisions on a consistent basis. If you think about the diversity of businesses, we serve over 700 different industries. Every industry is very different in the way it operates. If you think about a restaurant, for example, it may have very distinctive cash flow and revenue patterns.
We need to have signals and we need to have data sources that will enable us to understand how that business is working. We’ve built this capability over the last five or six years to be able to ingest data from multiple variety of data sources, transform them, and then aggregate that information so that they’re available for our machine learning platform. We have built a platform that we call Grover that takes all of this information over a hundred different data sources, and organizes them so that we can build models that are truly testable.