Artificial intelligence and machine learning technologies are proliferating in the alternative lending niche, but in late 2013, a unique concept saw its birth when Emmanuel Marot shifted from working with a photo sharing app and started working on an algorithm to assist him with his personal investing on Lending Club.
“One of my obsessions is to think about algorithms that can help with human decisions and make things simpler and faster,” Marot said.
Marketplace lending (MPL) began under the moniker “peer-to-peer.” The idea was for individuals to invest in each other’s ideas and bypass traditional financial institutions to fund projects and businesses. Somewhere along the way, however, the institutions got involved and “peer-to-peer” transitioned into MPL. As a result of this shift in the market, investing became complicated and time consuming for individuals. That’s because big financial players were using scripts and applications to identify loans and snagging up the best investments before smaller players like Marot could act on them.
“It took just two minutes to see new loans,” he said, “but a lot of the loans were gone in five minutes.”
Tired of being left in the dust by those who were automating their investments, Marot decided to create his own script to combat this disadvantage. After polling about 70 other investors in the marketplace to gauge interest, he decided to buy a domain name and build a website then invited other investors to try it out. That’s when LendingRobot was born.
What LendingRobot Does For Whom
After trying out the script for the first time, one of Marot’s early customers sent an email that read, “This is brilliant.” He immediately tossed $200,000 into his Lending Club account for LendingRobot to manage (managing MPL accounts at Lending Club and other platforms is its core business). Not long after that, LendingRobot announced they were going to become alternative lending advisors and Marot set about improving his algorithm. But let’s back up a bit.
At the beginning of 2014, LendingRobot collected $700,000 in a seed round and followed that in January 2015 with a $3 million Series A funding round. This allowed him to expand his service to meet the needs of investors using other lending platforms. LendingRobot now manages portfolios at Prosper and Funding Circle, as well.
“We don’t have custody of the money,” Marot said. “The client wires their money directly to the platform and we connect to the platform on their behalf. We look at how much the customer has available and submit orders on their behalf.”
There’s no charge for accounts under $5,000, but for investors with larger amounts, the fees are minimal.
“If you have $50,000 at Lending Club,” Marot said, “we’ll charge about $200 per year.”
Those management fees are way less than traditional money managers charge for Wall Street investments, which adds to the attractiveness of marketplace lending for many investors. Recently, LendingRobot rolled out a new service called LendingRobot Series, the first robotic hedge fund in the alternative lending niche.
“The client wires their money to us and we manage it as a fund by investing in the origination platform,” Marot said. “It’s a more complex product and there are a lot of fees we have to pay, so it’s more expensive for the investor.” This vehicle costs investors about 1% in management fees.
The Technology Behind the Robot
For the robo-fund, LendingRobot added a platform—Lending Home. This allows investors to diversify their MPL portfolios.
“It’s not hard to connect to a new platform,” Marot said. “But very few of them have APIs to allow us to do that. We could do it by scraping, but that’s ugly and fragile.”
Marot said that was how he started with managing Lending Club accounts for LendingRobot customers, but due to safety concerns and the risks involved, he decided to use APIs instead.
Rather than work off-the-shelf, LendingRobot writes its own algorithms, but they are based on well-known techniques.
“We settled on computational statistical analysis (CSA),” he said, “which is useful in medical studies. If a patient has an illness and you want to treat it with medicine to increase survival chances, how do you factor that in? We found that process worked well with the products we were offering on loans. You have a loan that starts paying, but you can’t guarantee that it will continue to pay, and that’s where CSA comes in.”
Because loans can be paid back, once the payoff is complete, the loan disappears. LendingRobot uses a mix of CSA and a machine learning algorithm to service and manage loans through the platforms it connects to on behalf of its customers.
LendingRobot’s machine learning algorithm is helpful in determining how to weigh all the factors that go into managing loans on different platforms for each client. It’s not important for people with low or sterile FICO scores, but it is instrumental in finding mathematical curves and balancing the weight of each of the factors involved—loan payment amounts and monthly income, for instance. They’re based on the time value of money calculations and the probability of a certain loan defaulting at specific times during the life of the loan. Using advanced mathematical calculations and logistic regression, LendingRobot can make certain predictions and use that knowledge to determine the value of its loans.
Keep It Simple, Robot
Marot aims for simplicity.
“We want a tool that normal people can use for their P2P investments,” he said. “They don’t want to look at numbers for hours. It should be a no-brainer to invest in P2P loans.”
Because LendingRobot was able to turn complex machine learning algorithms into a simple platform that makes investing easy, the company has seen 400% growth in terms of assets managed year over year. But Marot believes they can be much bigger.
“It takes time for people to become familiar with new asset classes,” he said. “When they do, alternative lending will become more popular.”
Marot is confident the alternative lending niche will grow as more people learn they can get better investments and pay fewer fees.
“People don’t need to see all the intricacies behind it,” he said. Clients can open their own accounts at Lending Club, Prosper, and Funding Circle and then open an account at LendingRobot to manage them. For Lending Club, the minimum investment is $25, but Marot recommends at least $5,000 because it allows an investor to invest in hundreds of loans. The more diversified the portfolio, the lower the risk.
“If you invest in 250 notes,” he said, “then you are more likely to get stable returns.”
LendingRobot’s smallest customer has $300 in their account. The firm’s largest client has $4 million, which means investors don’t have to be accredited. Anyone can get in.
Marot narrows the demographics of his clients into two main groups. The first group are older investors who aren’t retired but are looking at retirement. They’re typically well-versed in investing but aren’t financial experts. They’re doctors, lawyers, and other non-finance professionals. The other group consists of younger people, early 30s usually, who aren’t rich but are doing well. They’re technologically savvy and understand the importance of investing, but they don’t trust banks or the stock market.
“Our clients expect transparency,” he said. “And simplicity.”
The Future of Lending
Currently, Marot said, the decision-making process for lending products is still manual. Lending Club still has people answering telephones for customer service, for instance. But Marot sees those tasks increasingly becoming more automated.
He also sees lending platforms moving toward specialization. It will become simpler to borrow money.
“Investors don’t care so much about where their money goes as long as it gets good returns,” he said.
Nevertheless, he sees the industry moving toward more specialization, with hundreds of platforms drilling down to specific niches of investment. Technology will be the facilitator of that specialization, he said. Already, it’s happening, because you can go to some platforms and invest only in real estate. Cloud-based operations allow companies to use services anywhere in the world, so companies can have their accounting department in Australia, customer service in India, marketing in California, and headquarters in New Jersey.
“It’s kind of crazy that it still takes a couple of days to wire money,” Marot said. LendingRobot uses Dwolla, a state-of-the-art technology company that makes money transfers easy. “In the future, we’ll have something more efficient.”
The blockchain, for instance, is a digital ledger that make smart contracts simple and efficient. LendingRobot uses the technology for its own contracts. Marot sees it becoming a bigger part of the global financial infrastructure. Because of the blockchain and its public nature, there’s no way for LendingRobot to be dishonest. It fosters transparency.
“We use a hash code system on our own ledger,” he said. “We can drill down to a single cent, show how many payments are made, know when a loan was paid off, and have a line for each asset in an account. We can calculate the hash code based on that ledger.”
LendingRobot is on the lookout for more platforms to manage for its clients, however, Marot doesn’t want to branch out into managing other asset classes. He plans to stick with marketplace loans. But to get to that final state of glory, they’re going to need more machine learning engineers and people who can work with algorithms.