It’s boom time for credit product startups. The last few years have seen hundreds of startups racing to make credit accessible and affordable. A multitude of factors are contributing to this growth - low interest rate environment, venture capital, mobile penetration, internet penetration, new financial/payments infrastructure, etc. This is true especially for developing countries.
There are many different types of credit products - Personal Loans, Mortgages, Buy Now Pay Later, Point of Sale loans, Auto Loans, Fix and Flip loans, Corporate Credit Cards, Business Loans, Credit Cards, Invoice Factoring, etc. but all of them can be understood better with this framework.
The framework has 5 key areas: - Insight - Data - Risk Model - Economics - Scale
Insight: What is the ‘new’ key insight?
This is the most important step before launching a credit startup. Clearly define what is missing or wrong with how things work right now. This insight will be the foundation for every decision you make and every action you take.
Examples: Stilt started with the insight that immigrants are mispriced by the traditional financial system. They are mispriced because not having a credit history is considered a negative. It shouldn’t be. Immigrants are financially responsible. That’s why they deserve affordable and low cost credit products. Brex: Startups can’t get credit cards without personal guarantee. But startups aren’t that risky. They should be able to get a card. This is because startups have raised money from VCs and that is a good indicator of risk that others aren’t using. Mortgage Startups: Founders are denied mortgages because they have variable salaries or they own their companies. But that shouldn’t be the case. Founders are lower risk because of their earnings and future potential. They should be eligible for traditional mortgages.
I have seen that some founders don’t spend enough time at this step. A truly differentiated insight comes from personal experience and/or deep analysis of the problem. I can’t emphasize enough the importance of being crystal clear on the specifics of the wrong that needs to be fixed.
Next, start collecting data to build risk models based on your insight. The data can be collected manually (in the early days) but they should map to your proposed way of fixing the problem. e.g. Let’s continue with Stilt - if we believe immigrants are lower risk, then we need alternative data to prove their low risk. Education, employment, and international data sources exist that can be used to underwrite immigrants. For Brex, using cash in a startup’s bank account to underwrite can be used to prove that startups with funding are not risky. In both examples, startups are using unique data to solve the problem of mispriced risk.
The data collection should be scalable in the long-term and it should be stable and consistent at scale. It’s not the most important concern but should be kept in mind when proving the initial hypothesis.
Risk Model (or risk guidelines): Can you build a risk model on this data?
Now you have an insight and the data to (hopefully) prove that insight. The next step is to build risk guidelines/model. These guidelines will define who you approve, for how much, and at what rate. There are a lot of things to figure out. The underwriting is always manual and fuzzy at the start. It’s difficult to determine cut-off values without quantitative performance data. i.e. if you develop a score, how would you decide a value for approval? These are difficult questions with no easy answers. The truth is that you have to test a combination of values to see what works.
That’s why it’s so important to actually start lending money. As real dollars are on the line, you may realize that you missed important factors or discard other factors that you thought were important. This is an evolving process. But to prove your underwriting, you need data based on actual loans.
Performance/Economics: Can unit economics work (in theory)?
Disburse a few loans as quickly as possible. The speed and volume of loans depend on your target market and your loan product. Compare payday loans to mortgages. Lending small amounts for a short time frame is simpler, while mortgages need large chunks of cash. The exact metrics to determine performance vary by the loan product, but early performance is important.
That’s why generating loan performance should be your next goal. Repayments and performance on a few loans work way better for fundraising than just a theoretical risk model.
The loan performance data help with refining the business model. If unit economics are really bad i.e. you lose more in defaults than you make in interest or net spread (interest - defaults) is zero, you’ll need better pricing or a new underwriting model. Loan economics is really important for credit startups because that’s the only way to make money. Models improve at scale because of more data but the scope of growth is a mystery. You will need to make that judgment call based on your target market and potential spread.
Generally, data and underwriting break at scale. This happens if your data or underwriting guidelines were too specific (overfit) for your initial population.
After proving your business model on a small scale, you will need to grow fast. To do this well, standardizing data, its collection, and underwriting become really important. In some cases, the data used for underwriting have to evolve. Growth also means a larger focus on compliance.
Credit is a highly regulated industry similar to insurance, healthcare, etc. As you grow, it is important to build regulatory compliance as a first class function. It can be argued that compliance should be the first step, but I believe it’s not important for startups if they are not blatantly violating regulations.
If data and underwriting models hold as you grow, credit products are highly scalable and predictable sources of revenue.
Hope this is helpful when thinking about your credit startup.
Feel free to follow me or send me a note on Twitter @rohitdotmittal. I am always around for discussions.
Please subscribe if you liked it and would like to receive future essays.