Affirm Holdings, Inc. (AFRM) Earnings Call Transcript & Summary

Thanks for joining us. I'm Jason Kupferberg, the payments and IT services analyst here at Bank of America, and we're very excited to have Michael Linford here, the CFO of Affirm. And we're going to jump into a bunch of fireside chat questions. We are going to give the audience a chance to ask some as well. So Michael, thanks for joining us. Appreciate it.

Thanks for having me.

Yes, plenty going on, lots to talk about. I'm actually not going to start with a credit or funding question because -- we'll get to those. Don't get me wrong. But let's actually start with competitive landscape. I would love to get your perspective on buy now, pay -- or how has the landscape evolved? How would you kind of segment it? We obviously saw the announcement that was long awaited from Apple about Apple Pay later. Where do you think that might fit in? And we can take it from there.

Yes. So I think the -- it's refreshing to return to the part of the conversation about, I think, is the most important thing for us, which is the fundamental shift in consumer behavior that we're tapping into. And we talked a lot about that over the past year, I think over the past 6 months with the uncertainty in the macro environment. We had some talking about it. And I think a lot of folks have forgotten what's actually going on underneath the surface, which is consumers are still voting with their wallets away from traditional payment methods, credit cards and debit cards, and finding new ways to extend purchasing power and to do it safely without the fear of revolving debt. And that trend is going on, has been going on for a few years. It's the core thing that we tap into, and we think it will continue. I mean Apple's announcement this week is another marker on that. It is a proof point. And this is not a nichey thing on the side. It has gone mainstream. And it's our view that this will continue to take share in terms of consumer spending and will continue to be a material part how consumers --- are things how merchants can effectively convert transactions. I think that the landscape is evolving, and it's not surprising. Anytime you have the kind of hyper growth this industry has had, where -- we estimated it's a little over 3% of GMV, e-comm GMV last year. Estimates range about 8% to 9% over the next 3 years. I think it will be slightly higher than that. Anytime you have that level of secular growth, you do have a lot of people who want to participate in the game. And you've seen that. Ourselves and a few of the kind of pure-play companies have been in markets selling these services to merchants and consumers over the past several years. You've also seen companies like PayPal enter the market last year. They obviously entered and generated a lot of volume very quickly. You've seen the credit card companies try to do a post-purchase, a post finance product that they want to think is the same idea of building an installment loan into the core credit product. And yes, it's going to continue. And think that Apple's announcement is really another piece on that. And I would say that we -- there's more we don't know about. Well, we do know Apple is, obviously, quite secretive. I wouldn't worry about trying to ask the question: Is Apple a lending company or a tech company? I don't think you have to answer that one. Because I do think that what they're trying to do is engage their consumer with a product, which I think is a really good thing. They raised their hand and joined us in the no-late fee version of the product. We think it's really important because we think you can create an honest financial product when you do that. It's important that you can underwrite well when you do that. You can't rely on carrots and sticks in order to get people to pay you back. You actually have to do the underwriting. And they joined the ranks of ourselves and our key partners like Amazon and Shopify in saying that these products are all without late fees. I think it's a good thing for the consumer. I think it's a good thing for the industry to have such a strong technology leader like Apple offer these products to their consumers. I think it'll be a really cool feature for Apple consumers. I don't think it impacts us as directly, because I think our predominant way we go to market is not through wallets. It's actually through merchants. And those merchant integrations are things that we spent a lot of time building over the years and those relations are very deep and very sticky. And our ability to command strong MDRs and multiple revenue streams is very different than what the Apple Pay product -- Pay Later product looks like today, we think. Again, there's more we don't know about it than what we know. So we're not super stressed about it. We think it's, on the balance, a good thing for us. And we're particularly proud that they're helping set the standard on no late fees. Meanwhile, I think what you're seeing in the pure play, the NPL space is a bit of a retreat. I think you're seeing folks be a lot more risk off, which is a total opportunity for us. I think we've always been a leader in how we think about risk, underwriting, capital markets, credit. Those are things that we actually invest a lot of energy in and we're quite good at. And as a result, in these environments, we're able to take share from our competitors. So I think we're a little bit more risk off given the uncertainty in the macro environment.

A lot of chatter in the market about the health of the lower-income consumer, and just wanted to get a sense from you. Now obviously, with stimulus being in the rearview mirror, can you give us some sense of the income level, ranges of the Affirm user base and where the kind of GMV concentration is by income level?

So we don't give detailed stats to everybody to be able to break it down. I think the thing -- directionally, we tell you it's very representative for the U.S. consumer. When you cover 60% of U.S. e-commerce with the likes of Amazon and Walmart, you end up with a pretty good representative sample of the U.S. consumer. There is a lot of chat on the low-end consumer, the low-income consumer. And I think there are just a couple of things that are important there. The first is we're really good at stacking risk in our business. And the way we actually measure whether or not macro environment has caused an impact in our ability to effectively price and underwrite loans is whether or not we see a drift in our models. And so as you can imagine, in this environment, we're quite diligently studying every possible data point to see if we have any drift. And today, we see none. And that's not to suggest that you won't see it. But we have a lot of confidence that even as things do deteriorate for any subset of consumers, our short-duration asset combined with our strong orientation in data and data science and a really good model that sorts risk quite well will allow us to generate high-quality assets in any environment. We included a slide for all the investors and then we included a slide in our last investor presentation that kind of gives you a sense of the shape of the curve. And it's by our credit score, not by income. But those things do connect. And if you look at it, the steepness is quite sharp. So for the first several deciles of consumers, the losses are high, but relatively flat. And then it's the last couple of deciles where you really do all of the -- the money is made, right? And in particular, the steepness is very, very, very sharp on that marginal consumer. So the last loan we make is the least profitable loan that we make. And that ability, that insight, that steep curve on the underwriting side allows us to be very selective about risk and dial that up and down. And that's not just a statement of confidence, which it is, but we did it. I mean that's what you saw in our results last quarter. The really strong results that we posted were a function of us really grabbing the steering wheel and being very thoughtful about the kind of risk that we're taking. And we saw the same thing during COVID. Admittedly, it was a flash crisis that was resolved with a lot of stimulus. But I think a lot of folks don't understand for that young -- the low end -- lower income consumer, that signals went away like the minute it hit. It did not have the same lingering effect that a lot of folks think it did. I think the bigger thing on people's minds now is the inflationary environment putting pressure on that consumer, more so than the lack of stimulus. And there, we haven't seen it. We actually think that the growth in nominal wages for our consumer puts us in a really strong position to help the consumers pay for the goods that are obviously more expensive in an inflationary environment, but with the nominal growth in wages, more than makes up for the risk that might exist there.

So maybe let's pull on that a little bit more. Your CEO, Max, had a blog post on Friday talking about how the loan book might perform in a downturn. So maybe just to drill into that a bit further. The message that management has consistently said is that Affirm can essentially control the level of delinquencies, right? And that, I think, gives you the confidence that the business will be resilient in a downturn. So maybe you can just talk through that a little bit more how you guys think about solving for a certain level of delinquencies. I don't know if there's any numbers you can share there. But what levers do you call? What are the signals that you monitor to make those determinations of what the optimal delinquency level might be at any point in time?

Yes. So the first thing is remember the duration of our asset. In the Split Pay side of our business, the asset is super short. It's a 3-week weighted average life, which just means that you're getting signals real time. And the book is a function of the front book at all times, if that makes sense. You're not carrying around a back book that you're trying to manage. It's the decision you made today that will impact where you are in a month. And so then it's about being hyper vigilant about where you take your approvals. And because what we do is different than some of our competitors. We actually underwrite every transaction. We do sort risk for every transaction. And so on the Split Pay side, we can pick the level of loss. And in 3 weeks, we get a very good signal about whether or not that level of loss is where we expected it to be. And if it is, we're continuing to operate there. And if it starts to deviate or drift, like I said, we would be able to tighten. And again, it's a function of the very large data set that we have. Last quarter, I think the number was 20 million transactions that we were processing on the platform kind of which we approved. That's a lot of addbacks. And it's a bit like a baseball batting average in that -- it's not that you need to hit a homerun every time because they aren't long 30-year mortgages. This is about the law of very large numbers, which means it blends itself quite well to a data science-driven approach to thinking about the problem. We don't have human underwriters evaluating credit files. We have models that help us estimate and sort the risk. On the longer duration part of our business, we do have to think about that. But again, a 5-month weighted average life still pales in comparison to most of the other folks who are used to lending in any cycle. And so the approach is very similar. We're just a lot more thoughtful about where we draw those lines. So we have an internal credit score called ITAC, which is a transaction level credit score. We don't score users. We score transactions. A transaction can be higher risk depending upon the channel the consumer is shopping in. It can be higher risk based upon the time of day. My favorite very small and unimportant example is that risk at 2:00 in the morning goes up, and you can see it clear as day. And it just goes like this. And then by 3:30, it goes back down because they're in bed again. And it's true. I'm not making this up. And so it's a small signal, but you can eke out a marginal benefit in underwriting by taking the whole transaction into account. The same consumer is a higher risk at a different time of the day. The same consumer is at higher risk depending upon what they buy. And we take all that information in and we make a risk calculus on that particular transaction. And we do that a lot. And then we learn from what worked and didn't work. We learned from new signals. And we're constantly rolling out new models. We're not public about that. We might start celebrating that a little bit more for folks when we roll out new credit models because they are really interesting. And the quality of some that we get is huge there. And so because we have all that data and because we have the discipline to say we're going to pick a risk point and decide where it is and then we monitor the heck out of it, and a short duration asset allows us to correct if things ever deviate, we have the strong sense that we're in control. I don't show these charts, but if I showed them to you, like I promise you, you guys would quickly roll your eyes back because there are like these hundreds of lines and cohortized charts. And we can do that because we're not looking at making a qualitative assessment anywhere. It is a data-driven approach and the data speaks to us. And if we see a deterioration, we react to it. And it's about the process to build the model and then the monitoring behind it. And those 2 things we've invested in from day 1, even when it wasn't cool, right? It wasn't cool to do that a year ago. We collect the last 4 digits of your social security number and -- for the longest time in our Split Pay product in order to connect all the way back to your credit bureau file. And that was a thing that put us at a big disadvantage because our competitors didn't. We found a workaround to be able to still pull credit data without that last 4 of your social security number and still can put all the underwriting that we want to get to parity. But we were solving the problem all along about how do we get the credit data in a transaction, how we underwrite that transaction when a lot of folks in the industry were just thinking late fees and then excluding you from the system was enough. And I think what you're seeing now is it's not enough. And when you do sort risk, you get differentiated outcomes.

So it sounds like you believe that your whole credit model is a real favorable differentiator versus the peer group out there.

It is a differentiator, and we believe that the -- there's 2 things. The model is good. We can stack it against FICO, we can stack it against any other credit model, and we sort risk better. But we're not unique in that. There are other credit models out there that sort better than the traditional models because the traditional models aren't good. That being said, we have that advantage and the structural advantage of being a transaction level underwriter, where you're looking at tens of millions of transactions. And because you're looking at lots of little transactions and you're sitting and making yes or no decisions every single time the consumer would like to buy something, you have a very different set of credit outcomes than if took that same credit model and applied it to, say, a credit card company, who has to extend to the consumer several years' worth of capacity to buy things. We're not doing that. That's not the decision we make when we extend credit.

So do you feel materially different about the quality of your loan portfolio today versus, let's say, the start of the calendar year?

Yes, I think it's materially better. I think that we have been very thoughtful about where we are at. We talked about the small optimizations we made in our last earnings call. And I think we're pretty vigilant right now. I think we are concerned like everybody else is. We read the news. We watch TV too. We have the same generalized external concerns. We don't see it in our data, but it does cause us to be thoughtful. And so yes, I feel better about it today than I did 6 months ago.

So it's basically demonstrating that ability to tweak as needed as macro conditions change and...

It's to grow really fast.

Yes. Not sacrificing top line in the process.

Exactly. And that's the other thing. A lot of folks say -- I got this feedback from an analyst, not as smart as you, Jason, who said that -- our answer was unsatisfactory on credit. They're wanting us to say something bad. And I apologize -- which is not a bad thing here. We don't have to stop growing in order to be prudent on risk. The market we're in is so growthful. We're going from small single-digit penetrations of e-com, we're on a path to getting towards over double-digit penetration. You don't have to be so reckless on credit to grow really fast. And that's the way we operated a year ago and 2 years ago when the market was avoiding it. And it's how we're going to operate now. It's really no different.

Right, right. I mean the other argument that we've kind of been making is that everyone's fixated on credit and those metrics are super important. But at the end of the day, you guys are managing the business also to a certain level of unit economics. And you've been consistent really since the IPO, saying, 3% to 4% in terms of gross profit as a percentage of GMV. You've been comfortably above that, right? And I actually remember last year on an earnings call, I asked you, I said, "If you're at like 5%, and is 3% to 4% too low?" And you said, "No, it's not too low." Right? And that's when everything felt great. And here we are getting closer to the 3% to 4% range. And that's how the model is supposed to work. I mean I thought it was interesting on the earnings call last month. You guys basically said that you're very comfortable that you can sustain the 3% to 4% range in fiscal '23. You're not giving official guidance yet. But I would just love to hear a little bit more what kind of underpins that confidence because there are so many pieces that go into the gross profit equation?

Yes. I think the profit equals revenue minus cost, that's a big insight there. Write that one down. So on the revenue side, we feel really good about our sources of revenue. We earn on Split Pay very high MDRs. On interest-bearing -- the interest-bearing content on the loan is substantially above our cost. And so we have the revenue content there. And then when you think about the cost of what's going on. Well, you have the credit environment. And we just kind of have been talking about it. We feel really confident about our ability to pick that. And then you have cost of funds. And a lot of folks are really focused on the higher cost of funding the business. We talked about this on -- 2 earnings calls ago, we wanted to give people a framework. We actually did the math for you. We laid it out. And I think most, again, folks don't quite grab that -- our liabilities are -- the way we fund the business are kind of committed and they're out further than the assets. So we are originating assets into existing funding models that we have a high degree of confidence in and certainty in the cost. We don't have a lot of uncertainty about how we will fund the business next quarter and the next 6 months or the next year. Far enough out, there's a little bit of ambiguity depending upon the growth. But in the near term, we're pretty confident, and most of that funding is on a fixed cost basis. And so we know the costs it will take to fund our business next year. We have a high degree of confidence there. We have a high degree of confidence on taking the level of credit losses. And so at that point in time, the units are really well protected. The other costs in there are things like servicing and even processing that are just very easy for us to predict and deliver leverage on. And so we have a lot of confidence. The bigger thing that's going on, the reason why -- if you ask me again -- even if we post a number above 4%, I'll still give you the same answer of 3% or 4%. I'll be very consistent on that. And the reason why it's not bigger is just the mix in our products, we're mixing into the Split Pay product. And the Split Pay product is a 4% to 5% revenue product that can't possibly be making 5% of margin because there's cost. And so the -- when you mix a 1% to 2% business and a 5% to 6% business, it kind of gets you to where you're at. And the reason why that number is going to keep coming down is because we are mixing differently. It's a constant refrain from us. It was a big topic of the revenue take rates over the past year when we were mixing out of Peloton. And it's a big topic going forward around mixing into -- Split Pay was over 20% of our business last quarter and it's a really important engagement tool for us. But it obviously is a lower margin on a percentage of GMV basis because it's a lower revenue take product as well. And so we feel very good about that. Business is at profitability. We love it. We're going to keep investing in it. When you combine those 2 things, you are going to continue to have that number drift downwards into that 3% to 4% range.

Okay. Yes. We're halfway through. Any questions from the audience? Anybody wants to ask anything? Let's start here, and then we'll come over here. Sorry. You got a mic coming.

Okay. So I think you've talked about 3 or 4 different tranches of business, right? You've got the Split Pay business, and then you've got short-term smaller transactions, right, and then they go up. If we just think about consumer spending and potentially some of the bigger ticket transactions coming down as discretionary income levels get tighter, how do you think about that in that mix -- in that margin mix?

Obviously, it will impact it if we mixed fundamentally differently. You saw that. You saw the Peloton business, which was a very good example of a larger considered purchase that obviously impacted the shape of the P&L. But I would disagree a little bit with the premise. I don't think that the need to finance discretionary purchases goes down in these times. Our product means more to consumers. And we have a lot of conviction around what that product will mean to consumer, what the financing will do for a consumer who still needs a coach and still needs a TV even though the stock market is down.

So it will be a share issue for you, right? Because potentially in the United States you'll see the number of discretionary transactions at those higher levels coming down, but you'll then -- you'll aim for a bigger share.

Yes, there's going to be less coach purchases. And I think the -- this is -- again, this is not just shooting from the hip. What we saw early in COVID -- and the history was obviously written in a way where it ended up not mattering. But go back to 2020 in March: everybody was concerned and the level of uncertainty was very widespread. And what you saw with respect to applications, whereas you saw consumers who were middle and up the credit stack applying for credit in times that they wouldn't. And we think the same thing is going to play out in a normal recession, where folks who are going to be mindful about the risk of unemployment, they're going to be mindful about their total cash consumption, being more open to paying for things over time, which helps us. That's a good thing for us.

Okay. So your -- you see enough visibility in being able to manage that mix. My last question is -- and maybe I misunderstood you, but I'm not clear on the funding costs being fixed, okay? So can you just clarify that?

We fund our business in 3 ways. We have warehouse financing, which is floating rate, that's floating. We utilize, I think, in the mid-30s of our warehouse capacity. It's not the predominant way we hold assets to maturity. We use those mostly to park assets until we can do one of the other 2 funding methods for them. The other 2 funding methods are forward flow. These are bilateral arrangements with committed capital levels, where we sell loans to counterparties on a whole loan basis. So we'll sell the loan to an insurance company. We onboarded an insurance company in May. And we'll sell loans. They buy the whole loan. They commit to an extended duration, let's call it -- 18, 24 months is what we usually talk about. And we negotiate a price. And actually, we negotiate a pricing table, because each loan has a grade and everything else. And we sell it to that loan buyer on a fixed price. And so rates going up and down doesn't change that price until we hit a renewal for that 24-month period. There are some of our forward flow agreements that have some floating pricing in it. There are some of our forward flow agreements that have episodic repricing. So I don't want to be -- I'm not painting it with universality here. But a lot of our capacity doesn't have that. And then we also fund the business with the securitization program. I'm very active in the ABS world. And those are all fixed rate borrowings. And again, those deals are committed and long in duration. And so we might do a 24-month securitization like we did last year when rates were really good. And we benefit from that today because these are revolving securitizations.

Okay. So they're fixed, but they'll gradually roll off?

Yes. Think of it as like -- I think we've done 9 or 10 securitization deals. We have a handful of forward flow partners. We have a handful of ware -- so there's like dozens of different funding arrangements and SPVs that we have. And so each one of these things has a staggered expiration. And as those roll off, we have to deal with it. But it's like, for 2% of our cost -- or 2% of our funding, we might have a 2% increase in cost. And we have to manage that, and we will. But we have a lot more visibility in the near term. I think the rising rate environment isn't troubling at all for us until it gets to a much higher level of rates. And if it does, even then, it's a problem for us 18 to 24 months out, not today.

What's that level? When does the model break in theory?

It's a really good question. We word it differently than when does it break. We talk about it is when do we start having to change things in order to make it work. I feel our numbers -- like, hey, look, if rates are at 10% on a sustained basis, we have to change a lot in our business. But if rates are at 4%, we feel pretty good.

Short term, at the short end of the curve?

Yes.

Yes. Okay.

When you say [indiscernible] consider change in [indiscernible], does that include going back to the merchant?

Yes.

And does it mean…

It includes revenue and cost initiatives. On the revenue side, it's repricing with merchants and with consumers. On the cost side, it's everything we can do to drive leverage in the business, leverage on the cost items, including payment processing and the like. And there's levers that we have out there that we haven't pulled, that we would pull in those environments.

A question in front, yes.

Yes. So just going back to Apple Pay later. I know a lot is still unfolding. But what would you say the customer overlap is between Apple Pay and maybe other buy now, pay laters or just Affirm in particular?

We don't know because it's not even live yet. So it's a really difficult thing to say. We can tell you how often consumers go to use our app and put our virtual card inside of an Apple Pay -- an Apple Wallet today. And that's a data point. And it's a -- a trivial amount of our volume goes through that mode today. I think that if you start where I started on who our consumer is -- and it's a widespread consumer; it's kind of a percent of America -- and I think that's where the overlap would eventually get to. But I think Apple does index higher income. It indexes to folks who have Chase Sapphire cards. And that's definitely not our target consumer. I think if you compare, though, to some of the other BNPLs out there, BNPLs players out there, I suspect it's pretty far away from them. I suspect that if you look at like the overlap between the Block consumer, whether it's Cash App or Afterpay, and you compare that to Apple Pay, I'm guessing that's much further away.

Yes. Do you think you could talk a little bit to your strategy around offline, specifically kind of debit+ and some of your distribution partnerships?

I think off-line for our category remains the most exciting piece of huge upside. We always talk a lot about BNPL penetration of e-commerce, and we've always measure it that way. And of course, that's ignoring 70% to 80% of commerce. And I think the opportunity there is enormous. And I would say the industry has not really solved it. I think digital payments offline has not really been solved. It's still a thing that lags quite a bit. So I think it remains a big opportunity. One way that you can solve that is build an app and do NFC and everything else. And we've done that. We do QR codes at Walmart. And these things all suffer from any number of limitations that cause it to not get the same traction you can get online. And I'm so excited for the debit card. The debit card, the plus card, takes an understood consumer form factor. You don't have to teach a consumer anything. You don't have to worry about whether or not the reader is working at the store. You don't have to worry about the right lighting in the store or the WiFi is -- you don't have to do any of that. It's a piece of plastic in your hand and you can chip it. And you can access all of the financial products we can deliver behind it. We think that's going to be a big opportunity for accessing offline spend. I think that it won't be a way to acquire new users early. It's the right way to reengage with consumers offline, which is probably the most profitable and thoughtful way to do that, at least initially. Eventually, we'll be able to think about debit+ delivered virtually. But again, I think the magic happens when you can just make peace with the fact that -- I may be very old school, but a piece of plastic is an understood form factor that's pretty universally accepted.

Maybe just to build on that, can you touch on the revenue model for debit+? How you guys will make money on that product? And do you think it moves the needle on fiscal '23 GMV?

We're not giving guidance for fiscal '23.

Fair enough.

And we haven't included it in our guidance yet because it's still very early. I think the -- what might get scaled -- the numbers get pretty exciting when you think about how top of wallet the card does become for our most active consumers. Almost all active consumers are debit users in their normal life. My favorite thing to do, and I encourage all of you to do it if you're researching the space, is to go park yourself at a grocery store, even high-end ones like Whole Foods -- go park yourself at the checkout counter and watch what cards people are using. I'm guessing a lot of people in this room are like -- me and you have a credit card that you shop at the end of the month. But a lot of American consumers will take their debit card out and they'll debit their transactions. And that's really a good indicator for our consumer. They think about debit as pay now, and they think about credit as pay over time. And that consumer doesn't have to start to make a decision about which card to pull out anymore. They can pull out the one card -- and that -- we think gives us a lot of confidence of that becoming a more top-of-wallet card. And therefore, it gets really exciting at scale. The revenue model is interchange on the debit transactions. And then we bring in any number of additional revenue sources over time. At launch, it will just be the interchange. Over time, we will be able to deliver interest-bearing loans to the card. And obviously, we're excited about what we can do for merchants and brands as well over time in the card.

And the fact that it links to the consumer's existing checking account makes it that much easier to...

Yes, I mentioned...

Adoption...

I mentioned some cost initiatives earlier about what we would do. And one of them is -- today, we have a mixture of ACH and debit on the payment processing side. One of the features of debit process that will obviously use ACH to sell the transactions, which take some costs out of the system, which then we can use to reinvest in some of the features for the consumer.

Is there another question in the audience? I don't know if I missed one. Okay. So I guess let's move on and talk about operating profits. You guys said on the latest earnings call you expect to deliver -- sustained, I think, is the word you used, sustained operating profitability starting at the end of fiscal '23. So a year from now, essentially. So maybe just talk about kind of what went into the decision to announce that to the market, the parts of the cost structure that have the most flexibility in case you need to adjust spending in the event of unforeseen top line softness to stick to that commitment starting a year or so from now?

Yes. I mean we chose our words carefully like we always do, because we have very good advisers and also because we want to make sure people refer back to it. Part of the reason we said sustained is because we were profitable on that basis last quarter. And not because that was a north star for us or some sort of imperative. That wasn't something that we did with the intentionality for the purpose of a reported number. And I think that shows you just how close we are. And I think we weren't getting credit for what we believe is a pretty strong underlying cash flow generation in the business that I don't think most folks understood. And so we really wanted to make sure the market understood that. And then we wanted to make sure that we committed to it in a way that was -- allowed folks to start to think about where the business was going to get to over the next several years in light of the current macro environment, where I do think investors need some comfort on how we think about the level of spend in the business. But if we thought it was unnatural, we wouldn't have said it. We believe that achieving that level of profitability is something that will happen naturally as the business scales. So just as a quick reminder. We talked about it as adjusted operating income, which for us is -- think about it as revenue less transaction costs on the margin side. And then you've got 3 buckets of fixed cost. You've got our sales and marketing costs, our G&A costs and our technology data and analytics costs. Sales and marketing is a line that we do expect to show a lot of leverage on. And that's just because it's -- so much of our business now is about scaling investments that we've made on our partner side and being mindful about that. We're penetrated to 60% of U.S. e-commerce now. That number will continue to go up, but it won't go up as fast as revenue. And on the G&A side, there's just G&A. It's supposed to be leveraged. Like that's it. We went public last year. Obviously, there's a lot of public company costs that were year-on-year growth. We feel like there's going to be a lot of leverage we can deliver there. And now tech and analytics is actually -- we will, already, keep investing in. The same earnings call we announced that, and a few weeks prior, we had acquired 100 engineers from a company to round up operations and fast. And that is -- like that's the posture we'll have. We're going to build great products to scale the business, and we're not taking our foot off the gas there. So we really expect that line to continue to grow quickly, although, again, not as fast as our growth in our unit economics or revenue less transaction costs. And so the other thing about it is we're not cutting away, we're not making anything unnatural, we're not sacrificing growth. We're just acknowledging what's going to happen anyway, and we're really scaling into that level of profitability.

Let's talk about Amazon for a minute. They went live with you guys around November, I think, shortly before Thanksgiving. So about 7 months in or so. How is the volumes -- how have the volumes trended versus what you guys had originally expected up to now? Maybe you can talk a little bit more broadly about the evolution of the -- kind of the relationship on the ground between the 2 companies?

Yes. I think even internally, we forget just how recent this is. Like we've not been live with Amazon and Shopify for a year yet. And I think so much of the excitement for us leading into next year and the long-term prospects are really scaling and making full benefit of the distribution we already have, let alone the new distribution that we keep adding like our partnership with Stripe that was announced earlier this month. And so I think we have a lot of excitement for continuing to scale that relationship. It did go live just before Black Friday, and it went live with one product. It went live with one product in a way that was very down funnel. For us, we talk about up funnel, down funnel. Up funnel is where you see it when you're browsing the product, display page. Down funnel all the way at checkout. You get a lot of goodness the more up funnel you get. Amazon is obviously very thoughtful about the consumer experience and the real estate they have up funnel. And so we're working with them to find ways to optimize the program. But I'm pleasantly surprised with the traction that we've had early, both in terms of our ability to print really strong unit economics for Affirm, but also deliver a lot of volume and consumers' satisfaction, happiness. We have to measure and monitor our star rating on Amazon. I think it's true that we're the top-rated payment method on Amazon.com. And that's something that's really important to us and Amazon around really creating loved consumer experiences, because what we do is special and really valued by our consumers. And we're glad that's showing up in how they look at it. That's obviously not a financial measure, but it's a good measure of the health of the program and certainly how Amazon would think about it, I think.

I wanted to touch on the regulatory backdrop too. Last year, the CFPB announced they were collecting data specifically on the Pay in 4, as you guys call it the split pay market. And that data was submitted. I think March is when it was due. So presumably, they're looking at that now. So we'd love your thoughts on what that may result in. And then more recently, we've heard some rumblings about the CFPB may be looking at AI-based credit decisioning models. And I would love to get your perspective there, if you've had any conversations with the CFPB about that? And just any -- what your concern level might or might not be there?

It's always dangerous to speculate about a very powerful and important regulator like the CFPB. So I'm going to stay away from trying to predict anything about what they're doing. What I will say is that we are -- we have always put the consumer first in our business and oftentimes to our detriment in thinking about how we build our business. We don't charge late fees. We think a product like deferred interest harms consumers, so we won't touch it. It's important to us to put truth and lending disclosures on every transaction because even split pay loans are loans because you give somebody money and they pay you back over time. It's a loan. And a lot of people don't want to call them that. And we took a view early that, that's what they were and we behave that way. We think that does serve us well -- or should serve us well. We think that, though -- I think that's a mistake to think that the CFPB's data request is tied to anything super near term. Although -- yes, we will not speculate on where they're at. What I would say is that I think that they are recognizing that consumers are adopting this product in very large numbers. So it's important as the -- their mandate -- it's important for them to understand what this product is. It's important for them to understand how it works. And it's important for them to identify areas where consumers could be harmed. And I think that's what they're doing. I don't know they have an agenda beyond -- beyond that, they may. But I think it's important -- I think they think it's important. And that assessment comes from the fact that, like we saw in the conversation, the fact that Apple is now offering this product, means that a regulator like the CFPB really needs to understand it.

All right. We've got to leave it there. We're out of time. Thank you very much. I appreciate, Michael. Thank you.