Upstart Loans with Interest Rates Calculated by AI
Table of contents
We’ve written before about how artificial intelligence is powering a “new credit score” which is based on 100s of data points that are not reflected in your traditional FICO credit score. The ability to assess your likelihood to pay back a loan is a win-win for everyone. The lender wins because they have fewer loans that default and the borrower wins because they (presumably) can get a better interest rate as a result. This value proposition has resulted in a whole slew of “AI fintech” startups emerging that want a piece of this action. One firm in particular, Upstart, is offering a peer-to-peer (P2P) lending model where you can lend money to people on their platform as well as borrow money.
Update 04/08/2019: Upstart has raised $50 million in new equity funding to fund its expansion plans. This brings the company’s total funding to $144.1 million to date.
The Company first unveiled their peer to peer (P2P) lending product in May of 2014 and since then they have originated over $700 million in loans with an average loan size of around $12,000:
The first thing we did was to try out their “instant interest rate” functionality and the results were underwhelming. The questionnaire honed in on things like educational attainment, salary, credit score, and years at present job. Our fictional applicant applied for a car loan with a credit score of +750, 10 years of consecutive employment, and a $150,000 per year salary. Note that at no time was any social security number provided so nothing could be confirmed by Upstart. Here’s what the results looked like:
We’re not really sure what to make of that double-digit interest rate for a used car loan but that seems to be normal for Upstart since they claim that their average borrower pays 12% for a loan. Here’s a look at what the profile of their average borrower:
The above picture may help to explain why our interest rate was so high for an auto loan (we checked Bankrate.com and a 36-month used car loan is currently 4.81%). It’s a bit hard to read but it says “stated intent by borrowers may not reflect actual use“. Upstart has no way to tell what we plan to use that money for so it charges us credit card rates (presumably). Upstart claims that they save borrowers around 26% on typical credit card interest rates which according to Bankrate.com are as follows:
While the interest rate for our car loan is not much to write home about, what we’re really interested here is the “peer-to-peer lending” aspects of the platform which allow accredited investors to lend money to people using these sophisticated AI algorithms. We signed up for the platform after which we were given some statistics about what sort of returns we can expect such as the below:
Whereas in peer-to-peer platforms like Lending Club you are allowed to select the loans you want to take part in for as little as $25, Upstart allocates loans to lenders randomly with a minimum requirement of $100. The minimum amount you need to invest to use the platform is $5,000 and that means that you have a 98.9% chance of achieving a rate of return greater than zero. You can also expect an average return of 5.8% using the platform.
Let’s compare these numbers to another P2P lender that does not use AI, Lending Club, which has now originated over $26 billion in loans. Lending Club claims that they save lenders 24% on average when compared to standard credit card interest rates which is a bit less than Upstart. Where they don’t seem to differ that much is in the returns they offer lenders which are as follows:
From a lenders point of view, it doesn’t appear that AI gives you that much value add. You get around the same rate and the same likelihood of a positive return. What lenders should also be concerned with though is the quality of loans being offered up. One metric to look at which tells us the effectiveness of AI in that respect is simply the percentage of loans issued that default. Fortunately, Upstart gives prospective lenders the raw data for every single one of their loans so it’s easy to produce tables like this one which show all loans to date and the number that have been charged off:
Now what we can do is actually compare these numbers to Lending Club. In order to make this as fair as possible, we looked at all loans originated by both providers for the years 2015 and 2016. We’ll also combine Upstart “A” loans with “AA” loans to make things easier. Here’s a comparison of the percentage of loans that defaulted in each category:
While Upstart managed to outperform for the “E” and “C” loans, both lenders were actually tied overall at 5.48% defaults until we gave them the extra decimal point at which time Lending Club nudged out Upstart slightly. With all things being equal though, why not choose the firm that’s using AI since they should be able to get better and better at originating loans over time?
Using AI to calculate a credit score needs to ultimately result in fewer defaults and better interest rates. We talked to Avant recently and they told us how they focus on “near prime” which they said was “a FICO score of between 600-700“. In the case of Upstart, it doesn’t look like they are targeting any particular type of borrower.
If you’re an investor in Upstart, you’re probably pretty excited to read an article on Bank Innovation last week that says more than 25% of loans processed on the platform are now fully automated. That’s incredible to think about. What happens when you can get that number up to 90%? At 90% automation you can just reject the leftover 10% and you should be running as lean as you can possibly get for a lending firm. Then hopefully lenders can get a better interest rate on funds they lend.
As for borrowers, it’s easy enough to see what interest rate you get from Upstart, do the same at Lending Club, and pick the best one. Always pay attention to the final APR that you are charged which includes the origination fee.
Become a premium member and get access to hundreds of premium articles, reports and additional content.
Nanalyze Premium is your comprehensive guide to investing in disruptive technologies. Read by the top investment banks, management consultancies, VCs, and research houses. Trusted by over 100,000 institutional and retail investors. Covering disruptive technologies for nearly two decades.