Sustainably differentiated lending

August 2020

I’ve long held the belief that, to build a lending business that is differentiated and sustainable, you need at least 3 of 4 things:

  • Pre-income servicing: a way for the lender to get paid synchronously with, or before the borrower.
  • Proprietary data that others don’t have, that helps predict:
    • (more useful) the borrower’s ability to pay, and
    • (less useful) the borrower’s willingness to pay,
  • Lower cost of borrower acquisition than competitors lending to that same borrower
  • Lower cost of capital than competitors lending to that same borrower. This enables the lender to
    • profitably lend to the same borrowers at lower interest rates than competitors
    • expand margins by lending to the same customers at par vs. competitors
I came around to this framework watching Square Capital from afar. From the start Square Capital had 3 of the 4 on the list (and from the outside looking in I think Stripe, Shopify, Amazon, and Paypal probably have (or have had) similar dynamics playing out).

Pre-Income Servicing

Because Square Capital lends primarily to Square merchants and can intercept the merchant’s credit card payment receipts, they have pre-income servicing; they get paid at the same time (or before) the merchant gets paid. In Capital's case, Square gets paid when the merchant gets paid, directly out of the merchant's payment card receipts. Consequently, instead of depending on the merchant’s willingness and ability to pay, Capital primarily depends on the merchant’s ability to pay; as long as the merchant remains in business and continues to process payments with Square, Capital continues to get paid back.

In Square’s case, this access to the borrowers cash flow has similar dynamics to collateralized lending. In collateralized lending you have recourse to the borrowers current liquidity; pre-income servicing gives you recourse to the borrowers future liquidity. There may be other types of pre-income servicing, but this is what I’ve observed so far, and I think it applies to Stripe and Paypal as well. The best corollary for this in the consumer realm is payroll services (in the case of W2 workers) and gig economy services (in the case of 1099 workers). I wouldn't be suprised if, in the coming years, Gusto, Zenefits and Rippling launch consumer lending products. There are probably meaningful regulatory hurdles to cross, but it doesn't feel impossible.

In the case of 1099 workers this is slightly more complex. I remember seeing an Uber loan product a while back . Gig economy platforms, unlike payroll systems, have the ability to actually improve the quality of a loan; that is, Uber could lend money to a driver, and then funnel more rides (or more lucrative rides) to that driver, thus speeding up their payback and improving loan quality. No idea if they're doing this, but it's a lever that purely financial platforms don't have, that gig economy platforms do have.

Proprietary data driven underwriting

Square Capital has access to a ton of high quality data about a merchant. The longer the merchant is on Square, the better the data, and it includes (but is not limited to); business location, payment velocity, customer demographics, customer spend across other Square merchants, cash flow, etc. It’s like extending a loan to a borrower whose entire history with you has been their loan application, and they’ve been applying for a long time.

In many ways underwriting a merchant (or any business) is underwriting their underlying customer quality. An advantage Capital has in addition to lots of data on the merchant, is lots of data on the merchant’s customers; their behavior across Square’s network of sellers, their various credit and debit cards, how frequently they visit that merchant, which other merchants they visit, and how likely they are to dispute a transaction. This type of data about a merchants ability to pay is relatively unique to merchant processors (such as Square/Stripe/Paypal); I can imagine other B2B2C platforms, given the right depth of end customer data, can develop similar quality lending programs.

One bit of nuance here; it is technically true that a 3rd party lender could oauth into Square or Stripe via APIs and get access to that merchant's transaction data, and use that data to underwrite. However, it's extremely unlikely that the true depth of data that Square or Stripe has about that merchant will ever be available via API; this is not necessarily malicious or defensive; it's just that APIs embed tradeoffs like any other product, and all the underlying systems that these companies use generate some data, the full breadth of which might not be available for a 3rd party developer to access. For Square and Stripe, that data included everything in the auth stream, every datapoint previously ingested to underwrite fraud, and the underlying DNA to interpret and act on those signals along the way. From this vantage point I think that for the same borrower, Stripe has something of an advantage over Shopify, both because fraud underwriting DNA will be more native to Stripe, and because Stripe has a greater depth of access to the borrowers financial data than Shopify. All that being said, no idea what they've agreed to share.

One trap that I see from time to time; just because your data is proprietary, does not mean it's predictive for underwriting purposes. Tools like FICO and credit reports have been used & seasoned over trillions of dollars of lending instruments over decades, so the bar for a proprietary dataset to be useful for underwriting is pretty high. In the Square Capital case (and again I think true for Stripe, Paypal, Amazon, and Shopify) Square wasn't founded with the intent to eventually lend money; the expertise and dataset happened to be highly applicable to lending. That's not going to be true for every proprietary dataset, and in large part its hard to know in advance if your dataset will improve underwriting.

There are probably ways to get data about a borrower's willingness to pay, but I think it is strictly less useful than data about their ability to pay. How willing you are to pay back a loan is irrelevant if you're insolvent.

In addition, proprietary data is necessary but not sufficient; just because you have different/proprietary data than other lenders, doesn’t mean your data is predictive of a borrower’s ability or willingness to pay.

Lower cost of borrower acquisition

Capital doesn’t have to acquire customers because Square’s already doing that upstream, for its Point of sale, payments, appointments and other product lines. This translates to higher margin per borrower relative to a lender that has to originally acquire the borrower themselves.

There are a few secondary consequences for lending products when the borrower has been acquired upstream. First, as the underlying product gets better, you’re likely to improve the borrower’s quality overall; if the tooling Square provides you improves your marketing, lowers your costs, or frees up bandwidth, you’re likely to be a better borrower than might have been indicated by the data upfront, or a better borrower than you might have been for the same loan amount, for the exact same business, on another platform. In addition, if those improvements are real, the underlying Square product is likely stickier for you, making you less likely to leave, and marginally more likely to pay.

Second, the few companies I can think of started as something else and bolted on lending; the original, underlying products are pretty strong all by themselves, and the teams recognized a working capital need and stepped in to fill it. Genuinely not sure how to interpret this, but my current read is that many high quality lending businesses in the future, are being built today as something else (Square & Stripe built as PSPs before going into lending, Shopify being built as a merchant services company, etc).

Lastly, when you already have a relationship with that potential borrower, you can create push based acquisition tactics. This means inviting an extremely targeted group of borrowers to apply for a loan, where all their details are already prefilled. This is in contrast with a 3rd party lender saying "send us your Square data and we'll underwrite you."

I think one reason many tech enabled lenders in the last decade have struggled is that many of them have at best 2 of the 4, and in many cases the 2 they have are not truly sustainable. Lending Club for example really only had the lower CAC advantage because they were one of the first to lend online. They used functionally the same data that Chase had access to (FICO & proof of income) and they got paid the same way Chase did. Even the lower CAC advantage wasn’t sustainable - every lender that aspires to scale today is lending online, and in many cases they’re crowding the same channels (paid Google and Facebook ads in a lot of cases).

The last leg: Lower cost of capital

The primary way to have a low cost of capital is to have deposits as your capital base, which removes 3rd party banks/investors and their requisite financial hurdle rates from your cost structure. Until a few months ago, this was the domain entirely of traditional banks, and no technology company or tech enabled lender had been able to compete on cost of capital. A few events over the last few months have changed this:

These changes are different paths to the same thing; the ability to collect deposits, either from consumers or businesses. Time will tell how these organizations execute with their charters, but, the ability to collect FDIC Insured deposits give them a pretty insane advantage in the lending domain vs. other technology enabled lenders without the cost of capital to differentiate. Obviously, the 8k+ banks in the US have had the cost of capital advantage for a long time; but a large majority of them have operated at subscale. Without question, Stripe, Square, Shopify, Amazon and Paypal (just using them as the leading examples) have the ability, ambition and distribution to operate lending at a scale rivaling Chase, Bank of America and Wells. It's just a matter of time.

Thanks to Ariana Poursartip, Roberto Medri, Charles Birnbaum, Justin Overdorff, Audrey Kim, Matt Janiga, David Green, Femi Omojola, & Mengxi Lu for reading this in draft form.