written by
Renjit Philip

Affirm: Cost side of the Business model

Fintech Business Model Affirm 6 min read , January 9, 2021

Carrying on from my last post on the revenue side of Buy-now-pay-later giant Affirm's business model, I want to cover the cost side of the equation for this business.

Now, I know that no one is interested in the cost/expense side of the P&L, especially in Silicon Valley startups funded by VCs! Still, it is instructive for those operating in other regions of the world to dive deep into this. I promise to make it a tad more exciting by adding in a few customer metrics!

Thanks to the good folks at Affirm and their IPO bankers, we now understand more of the Fintech’s costs. Let us start with the Profit walk for Affirm as reported in their IPO filings.

Profit walk?

Profit Walk for Affirm
Affirm main cost heads as a percentage of their Gross merchant volume

The walk suggests that it a profitable business, but I would reserve that judgment for a few more paragraphs. What we need is a view of the profit & loss statement of the business to show us what is really happening under the hood.

The percentage split of the costs is as follows in the image below, and you can see that it is still a loss-making business. Net Income is negative.

Affirm cost as a percentage of Revenue

The losses have reduced over the years, and that is a good sign (Net Income losses have reduced as a percentage of revenue). It has been over eight years since Affirm was founded, and the business has not turned profitable yet. Fintech CEOs will have you note that unicorn fintechs like Affirm are not built in a day and that customer acquisition is expensive). Now let us shift our attention to the P&L statement.

P&L Statement

Affirm’s P&L

The most significant cost items:

1) Loss on loan purchase commitment:

What is this? I had written in my previous post that Affirm originates loans through its banking partner (Cross River Bank, NJ). Affirm then purchases the loan back, and under the terms of their agreement, they pay the principal, accrued interest, and the origination fee to bank.

In a few cases, the originating bank issues loans to customers that are zero or below-market interest rates. In these scenarios, Affirm incurs a loss by paying above-market rates (fair value of the loans), to purchase it from the bank. Advantage of doing this? Affirm does not have to be subjected to onerous federal and state regulations; this is taken care of by the bank.

2) Provision for credit losses:

This bucket of expense is a provision that changes according to the economic environment, the customer segment, historical charge-off numbers, and other factors. Let's take a look at the decisioning model curve.

This curve indicates Affirm's credit decisioning model, which looks like it is more accommodating than a standard FICO model. The decile's height is inversely proportional to the approval rate of customers in that decile. The higher is the curve from the X-axis, the worse is the approval rate. So, if you are Affirm and you use the blue curve to underwrite customers, then your approval rate should be higher that a bank using a FICO curve (orange color).

But the question is, if you are approving more customers, are you getting hit with higher losses? The graph below answers that question.

Charge-off curves

If you see the delinquency /charge-off curves above, it tapers off at 4% for 32+ months. It is too early to say if the machine learning models are improving or not, but it certainly appears that the later cohorts (Jan 19/ Jan 20) have a lower delinquency rate. This lower rate may be due to the tighter underwriting standards as the business gets more of an understanding of the buy-now-pay-later customer segment.

3) Technology and Data analysis costs:

Over the years Affirm has brought the technology costs down to 19% of its revenue. $122 million in costs is not too high when compared to the billions spent by incumbent banks. This cost bucket is where the credit decisioning models and the data management reside. The bulk of the IP that drives the valuation of the company is in this bucket.

4) Sales and marketing costs: Affirm achieved a coup of sorts by tying up with Shopify International in July 2020. It gives them access to millions of online merchants operating using the Shopify platform. In return, Affirm had to issue warrants that would accrue to Shopify over four years. The expense of the warrant purchase is allocated as an amortized sales and marketing expense to the P&L. This is an often-used tool to unlock strategic partnerships by well-funded startups such as Affirm. Possibly one of the few assets that they have on the balance sheet is the value of their unlisted equity!

What is the impact of Covid19 on this business?

Covid19 impact

There seems to be an increase in the dollar value of buy-now-pay-later transactions processed compared to 2019, and the delinquency rate is coming down. These are very early signals, and again could be a sign of the increased online purchases at the Affirm's retailer partners. Also, it shows that tighter underwriting standards were being applied in 2020. Now let us do some light number crunching to understand how these numbers operate on a per customer basis.

Light Number Crunching:

Affirm’s KPIs

So the math on losses per customer is that Affirm makes an annual operating loss of $30 on each customer. Why is that loss a down payment on the future of the company? The more loans that Affirm writes, the better its machine learning models become. Affirm has a ton of data on:

  1. Credit worthiness of its customers
  2. Transaction data (SKU/Cost/Numbers/Time/location etc.)
  3. Merchant sales
  4. Consumer demographics & purchase behavior
  5. Fraud signals

Future revenue sources

So, what is the point of incurring losses to gain all this data? Perhaps, the investors are playing a long game here, hoping that Affirm could become as big as a payment services provider (thus earning Merchant fees for perpetuity) or a lending institution (earning on interest rate spreads). Here are a few of the future revenue sources for Affirm:

1) A better credit decisioning model that FICO and possibly become a valuable data source for future Fintechs that are trying to underwrite Millennials and Gen-Zs

2) An opportunity to capture a larger share of the merchant transactions as Affirm brings more customers to their partner merchants.

3) SKU level purchase data can generate valuable insights for their merchants and the brands that they sell.

I know this was supposed to be about costs, but I did not want to close this without speaking to the opportunity generated by the spending.

In case you liked this, then you may enjoy the following:

If you liked this post, then you may enjoy reading:

  1. Fintech Valuation (inflated or not?)
  2. Blockchain: Basic building Blocks
  3. So, you want to start a Fintech?

Interested in managing your company’s content hub in a smooth and seamless manner? You can collaborate on content, schedule social posts, get analytics and multi-channel content distribution. Write once, publish everywhere. Their multiple integrations are time saving and make connections a breeze.

Look no further than Storychief.io. We have been using Storychief.io for years and have been blown away by the ease of use of the platform and the support. Give it a try!

Fintech Business model Affirm