Lending is as old as money and Embedded lending is the meme of the day. In this post, I unpack my thinking and thesis on where embedded lending needs to end up. But first, always start with the basics.
What are the fundamental building blocks?
The core components of a lending business are Demand, Supply, Underwriting/pricing, Loan servicing, Payments and money management, and Legal and compliance.
Demand and supply: In our economic model capital is the grease that drives the entire economy – it is one of the required inputs for the economic machine. The need for access to capital is the demand side of the lending business. To satisfy this demand for capital, you need a supply of capital to lend to the borrower.
Underwriting: The demand for capital and the supply of capital makes a market and the clearing price for this market is the price of the capital i.e the interest rate. This price is set by the underwriting function. Its core role is to set the amount if any should the borrower be able to obtain (loan amount) and at what price (interest rate).
Legal and compliance: As lending deals with money and money is always regulated :), everything that a lender does has to compliant with local laws and regulations. Everything from fraud to KYC to AML and maximum interest rate chargeable to anti-discrimination laws are in this bucket.
Loan Servicing: The capital borrowed by the borrower has to be repaid over some time. Sometimes borrowers fall behind on their payments or stop making payments altogether. Managing the entire process of calculating the re-payments, managing the collections process, and managing the delinquency and default activities fall into the loan servicing bucket.
Payments and money management: This is the core money movement infrastructure. Actual dollars and cents have to be moved through the banking system. Every cent needs to get where it needs to be and accounted for. Everything related to financial operations is in this bucket.
What are the major drivers of cost and growth?
Increasing demand (borrowers) and supply (capital to lend) are the major drivers of growth. On the capital side, it’s all about the cost of capital. Lending could also be thought of as the mechanism to determine the cost of capital. Cost of capital is defined as the interest rate the borrower is willing to pay for borrowing the capital. This is the same as the required rate of return that the capital holder demands to give up their use of the capital – rate of return and borrower interest rate are two sides of the same coin. Thus the biggest costs in a lending business are the cost to acquire borrowers on the demand side and the cost of capital to acquire capital on the supply side.
Modern Lending Version 1.0, the bank.
In the beginning, there was a bank. They sourced demand via their branch network. The borrowers who wanted capital literally walked in through the door. For the supply of money to lend, they relied on customer deposits. These deposits were sourced through savers who also walked through the front door. The customer deposits were assets on the bank balance sheet, hence the term balance sheet lender. As capitalism flourished and more importantly capital markets became sophisticated, the supply side of the equation evolved to include market-based financing options such as securitization. The model was still balance sheet based lending, except for the fact that the balance sheet portion of the bank could be expanded without taking in more customer deposits.
Banks could expand their lending operation by obtaining additional capital to lend via the private capital markets. The major growth driver for banks was their branch network. The bigger the branch network the bigger the customer base for both borrowers and depositors. In the pure balance sheet model, depositors’ expectation of required returns closely tracks the fed funds rate. Depositors value the safety of their money more than getting a huge return on their cash. This keeps the cost of capital for banks pretty close to the natural cost of money in the economy i.e the fed funds rate. This is the cheapest cost of capital available in the system! Banks monetize lending by earning the spread between what they charge the borrower and the cost of capital. So bank balance sheet lending has the best spread since their cost of capital is the lowest.
Rise of the marketplace, Lending 2.0
The internet opened up a new and cheaper (than owning branches) way to acquire borrowers (demand) and investors (supply) and this core insight powered the online lending movement. Constrained access to credit post the GFC also gave these upstarts a leg up in demand gen. These platforms directly matched investors and borrowers via a matching engine. They monetized via charging the borrowers an origination fee and investors a servicing fee. These platforms did not monetize the interest spread – they let investors capture all of it. This model is commonly known as the p2p marketplace lending model. Over time as these platforms gained scale, just like banks, they also took advantage of the private capital markets and started using securitization as a major source for funding loans.
This marketplace model attracted a ton of venture funding and a few of them such as Lending club, OnDeck and FundingCircle made it to the public markets. However, the returns of this entire vintage have been far below expectations. The hope was that these businesses would fetch tech multiples, but instead fetched multiples that were bank-like or worse. Where did all go wrong?
The causes were two-fold. On the demand side, customer acquisition became too expensive via the direct channel. Online advertising costs have only gone in one direction, up, especially with the duopoly of Google and Facebook. On the supply side, the cost of capital severely constrained the ability to grow the lending book. To amortize the high cost of acquisition these platforms had to increase the revenue and LTV from each user. This led to acquiring borrowers who needed larger loans and were ok with paying higher and higher interest rates. This led to a natural pull towards making riskier loans. As the loans got riskier the overall performance of the loan portfolio started to suffer (higher defaults) which led to investors with a low threshold for risk to leave the platform. To keep the machine going the platform had to go after capital which demanded higher rates of return i.e the cost of capital kept creeping up. Additionally, these platforms had made an explicit choice to not be regulated as banks and thus could not hoover up deposits as a cheap source of funding. This choice made sense as getting a bank charter is an arduous process. Thus the platforms got squeezed on both sides, with a higher and higher cost of capital on one side and higher costs for borrower acquisition on the other side. This became a huge constraint on growth and growth started slowing down.
The valuations were based on expectations that these businesses would be like software businesses i.e unconstrained growth with zero marginal costs – expectations that were not met in practice. These businesses were just internet versions of the banking business model. The street caught on and valued them like banks.
Baby steps into the future, Embedded lead gen, Lending 2.5
A borrower borrowing capital is a low-frequency transaction. They are not in the capital markets every day/month and are not looking to regularly engage with their capital provider. This infrequent transactional nature of the relationship is the core reason why acquiring a borrower via direct acquisition is hard and expensive.
Embedded leadgen posits that instead of directly marketing to borrowers it is better to go where the borrowers already congregate regularly. Examples for consumers are platforms such as Facebook and Google and for businesses, it’s platforms like Shopify, Amazon and Square. These platforms have aggregated customer demand by focusing on their core, recurring problems and providing them with superior products for those recurring workflows. As a result, they have an engaged and sticky user base that they can reach at zero cost. These platforms want their users to perpetually be in their walled garden – they want them to be forever sticky. Providing financial services to these users is a great mechanism to accomplish this. Why ever leave our platform when you can get everything here? However, offering financial products is a different beast with a ton of complexity and regulation – areas that these platforms would rather not get into. They would rather partner with somebody who does it for them behind the scenes.
However, this partnership is not like typical lead gen partnerships. The only goal these platforms have is to to keep their user in their walled garden. So the standard lead referral partnerships will not/do not work. For example, bouncing a user from Facebook to an online lender’s site to fill out a loan application will not/does not work. The loan product has to be offered within the context of the Facebook application – the user should never leave the Facebook app. Thus these partnerships need to be truly embedded in the platform’s workflow i.e these are deep API integrations. Embedded lead gen companies will be similar to API infrastructure companies and follow the same playbook.
A brand new hope, Embedded lending, Lending 3.0
Lending 2.5 solves the distribution problem a little bit but still doesn’t solve the entire problem for all the parties in the ecosystem. Even if you have deep integrations on the demand side you are still stuck with the cost of capital issue on the capital side. One way to solve this is to become a bank and take customer deposits. A lot of fintechs are going down this route – however, this is an extremely hard and cumbersome process. Additionally, any fintech that has to operate under a banking license regulatory regime runs the risk of being viewed/valued as an incumbent bank.
The other option for fintechs is to use private capital providers. Private lenders have only one mandate, earn a return on their capital above and beyond a hurdle rate. When private capital and fintech lending platforms meet the conversation goes something like this,
Fintech Lender: We have great underwriting, our target returns are 8-10% post-default. Here look at the data for all the loans we have already originated (also colloquially referred to as loan tape)
Capital provider: Great, let us take a look at the data and get back to you
Capital provider: Ok we believe you, but we still need a buffer – because we kinda still don’t trust you completely. So you need to add some buffers. These buffers are usually a combination of higher return thresholds (we need 10% return minimum), first loss provisions (you take the first 2% of losses), and concentration limits (minimum FICO, industry, etc).
Fintech Lender : <After some negotiation>. Deal. We will send you monthly loan tapes for ongoing monitoring.
So the cost of capital for fintech lenders is directly tied to underwriting success! If the underwriting is sound and keeps getting better the buffer reduces and the cost of capital keeps coming down. If the underwriting fluctuates the cost of capital increases as the buffer increases or the capital provider just walks away. Private capital is extremely flightly compared to direct customer deposits.
The reasoning behind the high expectations for the marketplace lenders was due to the (mistaken) assumption that they had an asset-light model. Hey, we are a marketplace connecting borrowers and investors, we are not taking any credit risk in this transaction! – we are a pure software business! However, investor returns were explicitly tied to underwriting and thus the marketplace platforms still had skin in the game and were tied to credit risk. The amount of credit risk didn’t really change relative to the balance sheet banking model in-fact you could argue that it actually increased. Balance sheet models had sticker customer deposits but marketplace lenders had flighty private capital i.e more risk.
To solve this problem, you have to decouple the pricing/underwriting from the platform. This can be done by providing a common infrastructure for capital providers to run their own models and decide the pricing for themselves aka Bring your own models (BYOM). This aligns interests across the ecosystem. The capital provider is best at assessing risk. They are the ones holding onto the risk (and reaping the return) and thus they should decide the price. The fintech platform excels at writing good software and should focus on just that. They should only think about aggregating demand and supply and providing the software infrastructure to handle the customer acquisition engine and back office functions.
Timing is everything 🙂 The AWS’fication of the world has finally made its way to financial services. There are a whole host of technology abstractions that are available today that simply weren’t available 5-10 years ago. It is now possible to build an end to end lending infrastructure platform as the basic building blocks are available behind an API. Need bank data → Plaid, need credit bureau data → Bloom credit, need money management →Modern Treasury.
Putting this all together, you have an incentive aligned model with each stakeholder getting the best solution for their needs. The end consumer gets financing at the place where they hang out the most and where it makes the most sense. The capital provider gets complete control and choice over the asset that they want to underwrite in their own way. They can outsource all the cumbersome back-end bits of legal/compliance managing the loan etc to the infrastructure platform. The infrastructure provider for all of this is building a true marketplace – it is connecting supply and demand and letting the market decide the price. It is a true software business!
This is the Embedded lending that we deserve!