Online lending marketplaces use software and pools of personal data to figure out who to lend to. Now a second layer of computers are picking through the choices made by the first.
MARK RALSTON / Getty Images
Most of the day, a small asset management firm in Seattle doesn't buy anything. A single computer just sits there, waiting.
And then, seven times a day, about 200 computers owned by Amazon light up and start replicating that single, sleepy server and begin running software to analyze loans made to individuals, largely to refinance their credit card debt. They're not looking for just any kind of borrower though — they want loans made to borrowers with prime credit, but not the best credit of the bunch, who usually want to refinance their credit card debt.
A package of those kind of loans that will earn an investor 10% a year. And then, once the analysis is done — about 850 milliseconds later — it's all over, and things revert to that single computer in Seattle.
This is the bread-and-butter business of LendingRobot, which manages its clients' money and invests it in loans issued by online lenders. The company started two years ago when two men whose companies had been bought by Microsoft, Emmanuel Marot and Gilad Golan, started buying up loans from Lending Club, the biggest online marketplace lender for consumers.
Soon, the pair realized the kind of loans they wanted to buy were being snapped up before they could get to them. Without automating the process, they couldn't diversify their holdings across many loans.
While Lending Club makes its loans available to investors to buy into over a two-week period, "in reality, the most popular loans are gone in five seconds," Marot said.
STEPHANIE PILICK / Getty Images
The other thing LendingRobot does is work on top of Lending Club's own secondary market for loans. Lending Club loans are paid back over terms as long as five years, and investors sometimes want to cash out before then, selling off their holdings to others. So LendingRobot dynamically prices its clients' loans until they're able to find a price to sell them at.
"If you want to put $50,000 in 2,000 loans, that's $25 dollars in each loan," Marot said. "The problem we're addressing is when you want to exit and cash out." Starting yesterday, LendingRobot can now automate the buying and selling of Lending Club loans.
Marot estimates that a portfolio of loans managed by LendingRobot will return 10% a year, while a random assortment of Lending Club loans would return 7%.
And he's an enthusiastic salesman for the product: during our interview, he twice asked I was invested in or wanted to invest in marketplace loans. And for his business to work, he needs to attract more money from clients. LendingRobot charges a .45% annual fee on the money clients put into the platform, with the first $5,000 free. This is a similar fee structure to automated investments for individuals in the stock market. Right now, LendingRobot has seven full time employees and $3 million in funding from Runa Capital with about $30 million in client assets.
Ultimately, Marot said, LendingRobot will be able to automatically buy up loans from a variety of online marketplace lenders that make very specific types of loans — real estate, small business loans, consumer loans for things like surgery or dental care — and create portfolios for its clients. "As a retail investor, maybe you want to put in some money, you don't want to open ten different accounts."
The irony, or the genius, depending on how you look at it, is that LendingRobot is using machine learning algorithims to automatically buy up loans whose underwriting is done is also done by code.
The core conceit of the rapidly growing marketplace lending industry is that software, scanning giant pools of useful personal data, can underwrite loans and assess risk better than old-school banks. Now, a secondary layer of software-driven intermediaries will pick through the loans approved by the first layer and make its own set of supposedly smarter choices.
It's robots all the way down.
No comments:
Post a Comment