Carvana Co (CVNA) 2021 Q3 法說會逐字稿

Good day, and welcome to the Carvana Third Quarter 2021 Earnings Conference Call. (Operator Instructions) Please note, this event is being recorded.

I would now like to turn the conference over to Mike Levin, Vice President of Investor Relations. Please go ahead.

Thank you, Matt. Good afternoon, ladies and gentlemen, and thank you for joining us on Carvana's Third Quarter 2021 Earnings Conference Call. Please note that this call will be simultaneously webcast on the Investor Relations section of the company's corporate website at investors.carvana.com. The third quarter shareholder letter is also posted on the IR website.

Joining me on the call today are Ernie Garcia, Chief Executive Officer; and Mark Jenkins, Chief Financial Officer.

Before we start, I would like to remind you that the following discussion contains forward-looking statements within the meaning of the federal securities laws, including, but not limited to, Carvana's market opportunities and future financial results that involve risks and uncertainties that may cause actual results to differ materially from those discussed here. A detailed discussion of the material factors that cause actual results to differ from forward-looking statements can be found in the Risk Factors section of Carvana's most recent Form 10-K and Form 10-Q. The forward-looking statements and risks in this conference call are based on current expectations as of today, and Carvana assumes no obligation to update or revise them, whether as a result of new developments or otherwise.

Unless otherwise noted on today's call, all comparisons are on a year-over-year basis. Our commentary today will include non-GAAP financial measures. Reconciliations between GAAP and non-GAAP metrics for our reported results can be found in our shareholder letter issued today, a copy of which can be found on our Investor Relations website.

And now with that said, I'd like to turn the call over to Ernie Garcia. Ernie?

Thanks, Mike, and thanks, everyone, for joining the call. The third quarter was another great quarter for Carvana as we continued our march toward becoming the largest and most profitable automotive retailer. We sold over 110,000 cars to customers, recorded nearly $3.5 billion of revenue, grew total gross profit by 100% and had our second consecutive quarter of positive EBITDA. Our team delivered these strong numbers despite entering the quarter with significant operational constraints facing the Delta COVID wave that peaked late in the quarter and despite making the choice to meter both retail and purchasing volume, which we implemented mid-quarter to ease pressure on the system while we operationally catch up to the demand we're seeing.

Since the onset of the pandemic, we found ourselves constrained in various parts of our operational chain. While those constraints have held us back relative to where we might have otherwise been, our team has done an exceptional job and has made tremendous progress throughout. In fact, in the third quarter, we bought and sold over 3x as many cars as we did in the third quarter of 2019 prior to the pandemic and roughly 8x as many as we did in the third quarter of 2018. This has all been possible because we have continued to focus on the long term. We continue to invest in the people, technology and infrastructure that are necessary to buy and sell millions of cars per year while delivering the exceptional customer experiences we've become known for. Given our opportunity, we believe this is clearly the right choice. To everyone across Carvana who has worked so hard to make our continual progress possible, thank you.

We've also recently entered into 2 exciting partnerships with Root and with Hertz that we discussed in more detail in our shareholder letter. While the potential of these opportunities is significant, they are both at their very early stages, so we plan to share additional details on our product, financial and scale ambitions with respect to these partnerships over time and will not be providing meaningful additional color at this time. These partnerships are each very exciting. Each has excellent underlying fundamentals and, therefore, excellent potential. But they're also exciting because of what they represent. As we continue building the Carvana platform to deliver exceptional customer experiences and to handle the constantly increasing scale, our horizontal and vertical opportunities continue to increase.

As has been the case since our inception, to maximize our opportunity, we must thoughtfully assess and continually increase the capacity of the business to effectively manage our growth and to simultaneously explore these opportunities. And then we must appropriately balance our ambition for both scale and scope with the focus necessary to make constant forward progress. So far, we'd like to think our team has done a good job of building more capacity in the business to tackle these opportunities and at managing these trade-offs given the capacity we have. To build Carvana into what it can be, we'll need to continue to spend energy getting this right in the future, and we plan to.

Looking forward, we remain extremely excited. We've reached the scale of over 100,000 cars and $3.5 billion of revenue per quarter, yet we're still only approximately 1% of the overall market. Our focus today is the exact same as it was at the very beginning. It remains on our customers, on the experiences we give them, on maximizing the value we provide to them, on leveraging technology to minimize our cost, on building for scale and on doing all of that with a long-term perspective so we can maximize our opportunity. The march continues.

Mark?

Thanks a lot, Ernie, and thank you all for joining us today. Q3 was another strong quarter for Carvana. Retail units sold totaled 111,949, an increase of 74%. Total revenue was $3.48 billion, an increase of 125%. Total gross profit per unit was $4,672, an increase of $616 and the second highest quarter in our history. Retail GPU was $1,769, a decrease of $88. The change in retail GPU was primarily driven by higher reconditioning costs, in part resulting from the impact of the Delta variant on production throughput, and higher wholesale acquisition prices, partially offset by higher customer-sourced ratio.

Wholesale GPU was $420, an increase of $154. This was driven by gross profit per wholesale unit of $936 and record wholesale unit volume, which grew 227% year-over-year due to buying more cars from customers. Other GPU was $2,483, an increase of $550. The increase in other GPU was primarily driven by strong finance execution and a positive impact of higher industry-wide vehicle prices on average loan size.

EBITDA margin was positive 0.2% in Q3. This marked our second consecutive quarter of both positive quarterly and trailing 12-month EBITDA despite significant investments for growth in 2022 and beyond. We ended the quarter with approximately $2.3 billion in total liquidity resources, giving us significant flexibility to execute our plan. We are executing well and remain focused on building our network and increasing our production capacity to meet demand.

We grew immediately available inventory to an average of 16,400 units in Q3, an increase from 12,800 units in Q2 despite the impact of the Delta variant on production volume. We remain on track to launch 8 new IRCs by the end of 2022 and continue to focus on growing our IRC teams in preparation for future growth.

The explosive growth in buying cars from customers we experienced in the past 2 quarters also placed significant constraints throughout our system in Q3. To ease the pressure on our system, we began metering both retail units and cars bought from customers mid-quarter to allow our operational capacity to catch up to demand. Most notably, to manage retail sales volume, we reduced the number of vehicles shown to customers in search results, which limited the benefits of higher immediately available inventory on retail units sold. We expect to increase our operational capacity in Q4 with an eye toward 2022.

Looking forward, we expect to complete a record year on retail units, revenue, total GPU and EBITDA margin in 2021. In Q4, we expect retail unit growth to continue to be governed primarily by our operational capacity. We expect revenue growth to be more closely aligned with retail unit growth in Q4 than it was in Q3. We expect total GPU in the low to mid-4,000s for the full year, marking our eighth consecutive year of substantial gains. We expect a seasonal pattern in total GPU in Q4, with Q4 lower than Q3. Finally, we plan to continue to invest in the business both to catch up with current demand and to prepare for growth in 2022 and beyond, leading to a seasonal pattern in SG&A per retail unit in Q4 and close to breakeven EBITDA margin for the full year.

We are extremely proud of the progress we have made as a company in 2021, navigating the unique macro environment while delivering rapid growth and managing through operational constraints. Our results relative to the industry continue to leave us more excited than ever about our long-term model and the path toward our goal of delivering more than 2 million retail units per year and becoming the largest and most profitable auto retailer.

Thank you for your attention. We'll now take questions.

(Operator Instructions) Our first question will come from Sharon Zackfia with William Blair.

I guess can you provide some more color on the metering that you're talking about, where you kind of took down the level of cars you were buying from customers and then the concurrent shift in the search results that consumers saw? I mean, how much of an impact was that to the quarter? Can you dimensionalize that? And then I guess, it sounds like you're hoping that will be kind of all back to normal in early '22. I just want to make sure I understood that commentary correctly.

Sure. So let me start with the constraints that we've been facing, and then we can kind of go from there. I think the primary driver of constraints was definitely the very significant growth we saw in the entire business, but most notably buying cars from customers in Q2, which then has continued and put additional strain on the business. I think it was accentuated by COVID, especially in the inspection centers. And then when we found ourselves a little bit behind, it was probably a little bit harder to catch up than it might have otherwise been given the unique hiring environment. And so I think that's kind of what's been going on from the constraint (inaudible).

To maybe make that a bit clearer, too, there can be increases in the total amount of work that are necessary inside the business that are actually greater than the increases in transactions that we see. And what I mean by that is if we get behind and it takes us a little bit longer to resolve a customer question and we have to call them back, we'll see more customer calls. If we have a delivery that gets delayed because the logistics network is kind of really full on certain legs, we have to kind of reschedule that delivery, and it can put additional strain on the logistics network. And so you can actually see kind of total work necessary in the business to grow more than just transactions. And so we saw that starting in Q2 and then spilling over to Q3.

Now let me talk about the nature of the constraints, I think, we faced throughout our life and maybe notably over the last 18 months and most specifically over the last 3 or 4. In general, demand sort of starts at the top the funnel. The customer shows up and is interested in buying a car, and the likelihood that, that customer chooses to buy a car from us is a function of many things. It includes whether or not they find the car they're looking for from us. It includes if they have a question, they want to call in and talk to someone, how long does it take to answer the phone or answer their question. If they're looking for getting that car delivered and they want to look at the delivery date, how quickly can we deliver that car.

And so when we say that we're constrained, what we generally mean is that kind of the amount of demand that we're seeing is causing those service levels to be below where we historically have had them. And therefore, fewer customers would elect to go through the process with us. And therefore, it curbs sales to some degree. And I think that's been active, as I said, throughout our life to varying degrees and certainly over the last 18 months. The difference this quarter was normally, that system can kind of balance itself out. And you'll see service levels kind of move out a little bit, and then that reduces the number of sales that you'll see conditional on demand. And then we catch up, and it kind of gets back into balance or at least start to approach balance. I think just with the speed that we saw total transactions grow, that moved more than we would have liked for it to have moved.

So we are in a position where we kind of chose to just take those same drivers, effectively a reduction of conversion, which there were several, but the most notable of which was choosing to be really surgical with what inventory was displayed in what places, and we just reduced the amount of inventory that customers could see in certain places. That ends up being a very powerful tool because we can reduce the amount of inventory that people see in certain markets if the logistics legs to that market are very constrained. We can reduce certain classifications of vehicles that maybe have more work associated with them on average if we have certain groups that we need to alleviate pressure on.

And so that was a very effective tool for us to use. And the effect from a customer perspective is just the average customer saw fewer cars than they would have otherwise seen. And therefore, they're likely to find the car they were looking for was lower.

In terms of quantifying it, wouldn't want to precisely quantify that. It was an effect -- it wasn't an overwhelming effect. I think the bigger effect is just kind of the degree to which we've been behind, in general, over a longer period of time. But that certainly was an effect in the quarter, and then that's the way that we implemented that tool set, to put ourselves in a better position, allow ourselves to catch up so that we can get back out in front of all the demand that we're seeing.

That's super helpful. I guess when you think about all this complexity, buying the cars from consumers and the weird environment that you're operating in, just in general, is this mainly a human issue at this point? Are you just behind where you want to be on staffing? Or is it a tech issue?

It's first order a person issue, and I think it extends to inside of Carvana, across all of the different operational groups and then also outside of Carvana. Many different groups that we deal -- when we buy cars from customers, for example, we're generally reaching out to a bank, and we're dealing with the payoff. And then we're asking for the title and return. Many banks are understaffed in this environment relative to where they'd like to be. Those processes can take longer. So I think it's first order a person issue inside of Carvana and outside of Carvana.

And then I think that across time, as we kind of outlined our priorities in the past, it's customer experience first, volume, GPU and then expenses. And expenses generally means building more technology to automate more of the process. So that's also part of it. Many of these processes, we can and continue to further automate across time, but that's generally longer lead time than the more immediate kind of fast-acting solution of getting more people.

Our next question will come from Zach Fadem with Wells Fargo.

So I just want to follow up a bit on the last couple of questions. Just with all the metering and Delta variant and labor constraints in the quarter, do you think the gap between your demand and actual unit sales widened or compressed in the quarter?

And then with your new IRC openings, is there anything you can share on timing or cadence of those IRCs as we move through 2022?

Sure. So on the first, I would say, we believe it widened in the quarter. And I think the kind of first way that we would articulate that is just in this quarter, we took additional steps to meter demand that we're proactive and we're just kind of natural system balancing. So I do think it widens.

Now look, I also want to make sure, whenever we're sitting here talking about constraints, I think I want to make sure and point out the success that we've had alleviating those constraints over time and the quality of work our team has done to do that. So I said this in my prepared remarks, but I got to repeat it. We've been constrained over the last 18 months. And yet over the previous 2 years, in terms of total transactions, cars bought and sold to customers, we've grown by 3x. And over the last 3 years, it's 8x.

So while we've been constrained relative to the demand that we're seeing, we've also been scaling very quickly. And I think the teams have done an exceptional job. They've been under a lot of pressure with the demand that we've been seeing, and I think they've just continually fought very hard to catch up. And so we're extremely grateful to them. And I do believe they've done a great job.

As it relates to the cadence of the inspection center rollout, we've only kind of provided the information that we expect to open 8 by the end of '22. And so we're going to stick with that for now for the kind of guidance that we're giving.

Got it. And then I can't let you off without asking about the Hertz partnership as well as the Root partnership, but so many elements to unpack. And I just wanted to drill in on the sourcing side as this appears to be a way of mixing in a higher source of new vehicles on your platform without having to go to auction. So just curious, do you think this partnership and partnerships like it expand your addressable market? And then on the unit economics side, to what extent do you believe your marketplace offering could be additive to the long-term 15% to 19% gross margin?

Sure. So let me start with this. I think in anything that we do, kind of independent if it's partnership or if it's something that we're building ourselves, generally, the framework that we used to evaluate it is we want to make sure that we can give a high-quality, differentiated customer experience. We want to make sure that we believe in the long-term economics of whatever we're doing, and then we want to make sure that we build it for scale because if those first 2 things are true, then you're going to want to be able to scale it over time. And I think that, that's the way that we generally think about all these opportunities.

I think from there, you oftentimes have this question of, do you partner or do you build something yourself? And I think there's a lot of considerations there. I think when you can partner, partners often tend to have assets that we might not have access to. You can book faster. There's sometimes questions about alignment that you have to try to resolve. You have to come up with terms that are good for both sides. So there's all kinds of questions around whether or not a partnership makes sense.

I'll start with Hertz. I think as it relates to Hertz, they've got an incredible asset, which is a high-quality flow of vehicles coming directly from manufacturers that have been utilized in their rental fleets and then ultimately make their way into the hands of another consumer. So they've got a really high-quality asset there. And then I think that would be very hard for us to replicate on our own. And then we would like to think at least that what we've built, which is a really high-quality, highly scalable platform to sell cars to customers, would be very hard for them to build. And so we're natural partners in that way. And so we're excited about the partnership.

I think from a scalability perspective, I think directionally, the way to think about these kinds of partnerships is it means that there's a little bit less work per transaction. And I think probably, that's the most important part of what scalable means. I think for something to be scalable, there has to be a lot of demand for it. And then the next question is how much kind of can you grow the amount of work that a company can do? And then to get to units, it's a function of how much work is there per transaction. So I think that partnership reduces the total amount of work that we have to do to get a car to a customer. And so in that way, it's exciting.

Now it's extremely early, and we've been testing this for a number of months. And then we signed a broader deal with them in the last couple of weeks. So this will take time for us to figure out and nail, but I think there's a lot of interesting potential there.

To try to touch on Root, I think we consider Root in the same way. The insurance business is a very complicated business. It's a business that has a lot more complication under the covers when you really kind of dig in that you might imagine from a distance, and I think that they've done a great job building a high-quality product over time. So we think that they're a really natural partner. We share a vision for what a solution is supposed to look like. We share a view that it takes a ton of work to build anything great, and I think that's really important because that means we can both pour work into it. And so I think that's also an exciting opportunity for us. But it's also very early.

And so as it relates to both those partnerships, we're very excited about their potential. But it's early, and there's a lot of work to do to get those to where we would like them to be. So we'll be working hard to try to get those to the place that we want.

Our next question will come from Brian Nagel with Oppenheimer.

So the first question I have, I know we discussed this a bit already. But just with regard to the operational constraints, maybe I just want to understand better. I mean, are you starting to see some easing at all in the various aspects of that? And recognizing that the environment is very fluid at this point, but if you look at it now, at what point do you think that -- would you believe that these constraints could let up and let the business flow much more naturally?

So that's hard, right? I mean, I would say constraints show up when we don't -- when we either don't accurately forecast the amount of work that needs to be done in the future or if we don't execute as well as we would like. That's what -- that's when they show up. And I think that when we look at where we -- I think that we've executed, I would like to think, at least, that we've executed very well through all of this. I think when we look at areas where we maybe didn't accurately forecast what was going to happen, I think several -- especially the first several COVID waves we didn't foresee, and that led to a decrease in efficiency and, therefore, constraints.

I think heading into the early part of this year in Q2, I think we did not foresee all of the growth that we would see, as I said, most specifically, in buying cars from customers. And so we weren't prepared for it. I think as long as we're prepared for it, we, again, would like to think that we've demonstrated the ability to execute to very high levels of kind of growth. And so I think that constraints really are about our ability to foresee what's going to happen and then our ability to plan for that and execute to that.

And so our continual hope is that this will be the last meaningful COVID wave. We'll find out if that's true or not. We don't quite know. The general kind of market across the board is in a bit of a unique spot right now. I'm not sure that was totally foreseen. And then as I said, buying cars from customers came faster than we would have otherwise thought.

So what I would say is we're clearly investing right now across the business as aggressively as we feel like we responsibly can to try to catch up because we do think there's a lot of excess demand there. And so it's hard to predict exactly how all this will unfold, but we're going to be working hard to grow the capacity of the business as quickly as we responsibly can.

Got it. And that's helpful. And then for my follow-up question, with regard to the efforts to buy cars directly from consumers, so you've been very successful. There's a number of other players now, both online and off-line, also pushing aggressively on this aspect of the business, so to say. So are you seeing greater competition? And does this -- does the fact that more players are now pushing this, does it ultimately have an impact the way the consumer reacts?

So I think that, that's the smarter place to err, and I think that's the right thing to assume. I think empirically, we're buying more cars from customers than ever, and we've seen a massive influx of demand in the last several quarters from an already very high level. So I think that, that speaks to the quality of the offering. But I also think that it's never smart to get comfortable and to stop building. And while we believe we've got the highest quality customer offering today, we also believe we can make it a lot better in a lot of really important ways. And we've got a great team that's focused on that product that has all kinds of interesting plans to continually make it better.

So I think undoubtedly, it's an area that's gotten more attention over the last several years. It probably will continue to get a lot more attention over the next several years, and we'll be continually building to make sure that we keep a spread between the quality of our offering and what else is out there.

Our next question will come from Rick Nelson with Stephens.

So I'd like to ask you about the ASPs. They've been on a rapid rise trajectory. Do you have any concerns or any about vehicle affordability and any potential pullback in demand given where pricing is today?

Sure. So I can take that one. So I do think higher used vehicle prices has an impact on demand through an affordability effect. Clearly, there are some customers that are targeting specific monthly payment that meets their budget. And as used vehicle prices rise, there are certain customers that are less likely to be able to meet that budget.

And so I do think if you look industry-wide, I mean, we're obviously growing incredibly quickly with 74% year-over-year growth despite the constraints that Ernie talked about. But if you look industry-wide, various data sources have total industry sales down, call it, on the order of 10% to 15% in Q3. And so I do think there's an effect there. We're growing really, really quickly despite that industry-wide effect, which I think is a testament to all the demand that we are seeing.

Okay. And I just have a follow-up. I'm curious about the unit growth that you were seeing during the quarter before you pulled things in or metered things. And I guess what was their final straw that caused you to pull back mid-quarter?

Sure. So I don't think we want to jump into the details of intra-quarter growth. What I would say is that what drove the decision was just that we weren't catching up. And I think that historically, we've been able to catch up faster, and we just saw that big influx of demand show up. And then we -- once you see that happen and once you're behind and then all sudden, you face a situation where the amount of work per transaction can start to go up, you run the risk of allowing yourself to get further behind, even if demand stays flat because the further behind you get, the more work per transaction there is, and that can cycle on itself. And so we decided to take the proactive measures to give ourselves a little bit of breathing room just so we could catch back up. And so that's what drove it.

And again, I think we feel really good about the progress that we're making there. We've given these stats of the growth we've made over the last several years. But we've -- even in the last several weeks, we continue to see improvement across all of those major groups, whether we're looking at customer care, who's answering customer questions and handling a lot of the transaction, we've seen our service levels get better there. If we're looking at logistics, we've seen that get better over the last several weeks. If we're looking at the inspection centers, I think they really have done an amazing job.

I mean, the inspection centers, that's hard. When you get behind in inspection centers, that's a big industrial undertaking. It takes time to catch back up, and they've caught up. They've built inventory in the quarter despite the fact that we faced another COVID wave, which is really tough in general. But it's extremely tough when you have an assembly line structure. So they've done a great job. Market ops, which handles last-mile delivery, has done a great job, and we've seen service levels improve there. So I do think we're seeing improvement, which is good, but we remain behind. And we're working hard to catch up.

Our next question will come from Rajat Gupta with JPMorgan.

Great. I just had a couple of questions on just the GPU component. On retail GPU, I guess just the quarter-over-quarter bridge, how should we think about the moving pieces? The intra-quarter price benefit was slightly lower versus than the second quarter. But was there any impact of higher recon labor costs or inefficiencies that impacted that number sequentially?

And just like given that pricing has continued to remain so strong in the fourth quarter, you talked about like a seasonal decline in the fourth quarter. So just curious as to why that would be the case.

And then just lastly, as you continue to face a big ramp of IRC next year and the labor environment remains challenging, I mean, should we anticipate any impacts to the GPU bridge next year and maybe just on wage inflation or just overhead expenses for the new IRCs? And I have a follow-up on F&I.

Sure. So let me take a stab at all of the questions in there. So first one, talking about Q3 retail GPU. So we had a strong quarter on retail GPU in Q3, came in at $1,769. I think if we think back prior to COVID, we'd normally expect to see some seasonal decline in retail GPU. I think we saw that for a couple of years prior to COVID. This year, we saw a slightly similar effect, even though the -- we're in obviously a very unique environment, where we did see early in the quarter, depreciation rates pick up a little bit relative to Q2. And so that definitely had an effect on the sequential change in GPU going from Q2 to Q3.

The other thing I'd call out is we did see higher reconditioning costs on a per unit basis in the quarter. That was driven in part by Delta, which definitely impacted production efficiency, which then increases your per unit cost. And so those impacts, we saw sort of a similar dynamic late in 2020. The impacts were smaller this time, but they were there and part of that sequential change.

Now as we look forward, you asked a couple of questions about retail GPU on a go-forward basis. So our assumptions for retail GPU in Q4 are embedded in our guidance for the full year. We typically don't break down individual components of that guidance, but we gave you some guidance on total GPU.

And then looking out a bit further toward 2022, you asked about the possible labor market effects on the reconditioning costs. And there, I would say it's too early to say. I think we can wait and see how that goes looking out in 2022. But in any event, any sort of labor cost adjustments are likely to be relatively small compared to the overall magnitude of retail GPU and overall reconditioning costs. So an area -- that would be my thought on your third question.

Got it. Great. Sorry, lumping a comment in one. But just on finance, in the third quarter or like the other GPU, the third quarter decline -- slight decline quarter-over-quarter, any color on if that was more finance-driven or service contracts? And just curious, broader picture, like how are customer delinquency assumptions looking for that business? Have they started to move back towards normal, albeit it could take many quarters? And if yes, I mean, how should we think about the trajectory going forward and the impact to just the other GPU?

Sure. Yes. So other GPU came in again at -- I believe we said $2,483. It was down a little bit quarter-over-quarter. I think the sequential movement in other GPU is well within the normal range that we would expect from quarter-to-quarter. So it's sort of well within sort of normal range of variability. In this particular case, I would say the biggest driver of the sequential change was just in finance, the -- some rate optimizations that led to slightly lower interest rates that came on sort of late Q2, early Q3 would be the largest driver of a relatively small change.

And then, Rajat, if I could also maybe provide a little bit of a perspective as well. I do think a useful exercise in kind of the way that we oftentimes think about the business and the way that we prioritize what we're doing internally, because I do think it can be very hard to predict exactly what's going to happen by line item by quarter. So we sometimes try to take a bigger view.

And so an exercise that I would suggest because I think if you look at all the different public retailers and you go back in time 20 years and you plot on a graph their EBITDA margin, I think what you'll find is an extreme amount of stability. There's very, very high levels of stability across all the different retailers, and very clear stories start to emerge. You can see some retailers performing a little bit better than other retailers over relatively long periods of time, which suggests relatively better execution. You can see what happens with the financial crisis. You can see what's happening right now. We're -- right now is a moment of interest in that 20-year graph, where you'll see that many of the franchise dealers are performing very well relative to history. And then you'll see that some of those that are completely reliant on auction are performing very poorly compared to history. And then those that have access to buying cars from customers are generally performing somewhat similarly compared to history, which is somewhat interesting.

But the reason I bring that up is because I do think that even going back to the reason that we started to report on total GPU when we first went public in 2017, the general idea there was just that there is a lot of ballots in this industry. There's tens of thousands of other dealers that are providing a fairly similar experience to customers, that have very similar cost structures to one another. And as a result, the industry is sort of structured in a way where some of the profits across those different components of the transaction tend to add up to cover their expenses and their cost of capital.

And once you get to that place where you kind of realize these lines have been very, very stable over a 20-year period, there is little squiggles kind of that you'll see in that chart from quarter-to-quarter, but they're probably not the most important thing. The most important thing is just what is the quality of the customer experience that you're offering compared to everyone else? What's the quality of your variable economics, both on the revenue and the cost side? And then if those things look good, how scalable are you?

And so when we think about our prioritization, that's the frame that we use, and that's how we try to make sure that we're putting our time and energy into the things that are most important. And I think through that frame, if you look at the last year or so, we would like to think at least that our performance looks very good. If you compare the way that we've performed to the industry in any important way, we think that it looks very good. And we think that, that's the most important thing because even when you're looking at kind of the bottom line, it tends to be -- there's a pretty stable return, in general, in buying cars from customers. And then there's some variation based on the way the business model is structured. And then it's about how you perform compared to everyone else. And we think in the long run, that's what really matters the most.

Our next question will come from Chris Bottiglieri with PNB (sic) [BNP] Paribas Exane.

I actually had a question. So it seems like you rebranded your third-party reconditioning units as marketplace units. At the same time, you're moving forward with Hertz as a large third-party seller. So if you think about your decision around the reconditioning units, is there an opportunity to use these partners in conjunction with third-party sellers like Hertz? Or do you foresee them reconditioning their own units? Maybe talk about that more, please.

So Chris, I apologize, I don't think we've really answered one of your questions in like 4 or 5 quarters. So we definitely owe you some answers at some point. But I do think it's early in the Hertz partnership. It's early in marketplace, in general. I think there's a lot of room for those choices to continue to evolve. And so I think we're not going to provide additional color at this time, and I apologize.

At least you're keeping it candid, I appreciate that. Let's move on to SG&A then. Is there a way to bifurcate -- further bifurcate SG&A cost between retail and wholesale? So I'm trying to understand like how much incremental SG&A you incur to wholesale a trade-in versus retailing it. Like further, if you buy a car from a customer but you don't sell one to a customer, like how different is the SG&A profile of that transaction versus buying and selling to the customer? Just trying to understand how trade-ins is impacting your SG&A per unit and personnel costs.

Sure. Yes. So I mean, I think if you start at a high level, buying cars from customers does incur additional operational requirements, and then it does incur additional costs. Some of those costs show up in cost of goods sold. That would be things like going out to the customer's door to pick up the car and bringing it back into our network either to be, yes, reconditioned to go up on the site or what have you. So that's one set of cost.

There's another set of cost that does impact SG&A. I think just one example of that might be customers calling in to talk about a transaction where they're selling their car to Carvana rather than buying a car from Carvana. So that's an example of a direct effect of buying cars from customers that does show up in SG&A.

There's also -- and Ernie sort of touched on some of these concepts, but there are also some indirect costs of the very rapid growth that we experienced in buying cars from customers in the last couple of quarters. And that just takes the form of everything just being a bit less efficient. And so I think there's both those direct costs and indirect costs that are driven by buying cars from customers.

Our next question will come from John Blackledge with Cowen.

This is Max on for John. I'm just wondering if you could talk us through the constraints on building in the pace of rollout for IRC build. You've already given the guidance for the end of the year. But just looking forward, what are the puts and takes on being able to roll out additional IRCs? And what's the balance there?

And then just wondering if you could talk us through a little bit, just inventory. Like is there a targeted inventory level that you'd like to get through? Typically, 4Q would be a build, but 1Q might be not building in a typical environment. Is that still true for next year? Or how should we think about that?

Sure. So I mean, I think we feel really good about our IRC trajectory. We expect to open 8 more new IRCs by the end of 2022. And then with -- when it comes to staffing those IRCs, we are making continual progress. I think many of the initiatives that we have kicked off and have implemented are paying dividends. And we're -- I think we're seeing nice progress there in terms of the second phase of scaling production capacity, which is building out the team. And so I think we feel good about all of that.

I think the -- on the inventory level, I think the simplest way to say it is we do want to build the selection of cars that are available to customers on our website. We made some progress there in Q3, climbing it to just over 16,000 immediately available units, up from just under 13,000 in Q2. But as we talked about, that effect was metered somewhat by the steps that we took to manage sales.

And so I think we want to keep building that immediately available inventory. We do think selection is a driver of conversion and a driver of sales and feel really good about continuing to march up that selection that we're making available to our customers, particularly as we make progress on relieving the other operational constraints throughout the system.

Our next question will come from Michael Baker with Davidson.

Two questions. I'll ask it at the same time, I guess, both related to GPU. One, does reducing the inventory that's available for customers to see that metering process, does that boost the GPU in any way in that you can show cars that have better margins because they don't need as much work or whatever the case may be? Does it give a lift to GPU? And then because of that, should we expect GPU to naturally decline?

And then I guess related to the GPU question, I think inherent in the idea that retail units will equal revenues is that ASPs flatten out. I think you've said that before, and of course, that didn't happen this quarter. Is there any specific reason why we should expect ASPs to flatten out in the fourth quarter? Or is that just sort of the way you plan the business and it may or may not happen? Or is there anything that you're specifically seeing?

Sure. Yes. So on the first question, the answer is no. The cars that get metered, sort of the algorithm that goes into deciding which cars to return in search results and which cars not to is not related to the particular GPUs on the car. So the answer to that first question is a straightforward no.

The -- on the second question, the -- in terms of ASPs, so I think if we think about revenue growth relative to retail unit growth, there's a couple of things that go into that. One is retail ASPs, and I do think that's a component. And I think when we give guidance on revenue growth relative to retail unit growth, that's one of the considerations that goes into it. Second consideration that goes into it is wholesale revenue will impact the ratio of total revenue to retail units, and that's another consideration as well. We did mention in Q3, we metered both retail sales volume and buying cars from customers. And then you might naturally expect metering buying cars from customers to have an impact on wholesale volume and revenue.

Our next question will come from Nick Jones with Citi.

Great. I think there's a kind of large franchise competitor that's getting into other adjacencies, like trying to, I guess, handle some of the logistics for powersports and maybe some other large heavy equipment-type logistics. I guess kind of what are your thoughts on kind of the competitive reaction? I mean, do you think kind of Carvana's success has caused some of the larger, maybe more profitable incumbents to react and maybe get ahead of some of the direction Carvana may be going in the future?

Sure. Okay. I didn't understand the first part of your question, but I think I got it now. So what I would say is, I mean, I think, listen, there's -- any time you're lucky enough to be successful, other people are going to see that. And so I think the kind of the default assumption anytime you're building a business and you're lucky enough to be on the right path should be to assume that others are going to notice and move in your direction. And then I think that our job as a business is to keep getting better. And I don't know really how much more color I can give you than that.

I think maybe one other thing that's notable and worthwhile would just be, this is hard, right? And I do think that it can -- I don't mean to imply that you're making this mistake, but I think from a distance, it can be easy to make the mistake that any given problem is easier than it actually is. And so this is a very fundamentally difficult problem. I'll start with talking about it on the used side, but you'll be able to quickly kind of adjust that for new products as well.

But when you're buying a car that is -- that varies in quality, when you're running it through a remanufacturing process, where you're putting $1,000 of parts and labor into that car, when you're shipping that car around the country in order to give customers a really broad selection, when you're handling trade-ins, when you're handling warranty, when you're providing customers with financing and you're verifying the information that they send over in financing, when you're registering cars across 50 states and all the different counties that vary in their registration requirements, there's just so much work that goes into that. And I think that it can be easy at first to start with kind of visibility from a distance of maybe the buttons that get clicked on a website and then to believe that, that's something that's relatively straightforward to replicate.

And I think our view would be that we worked really, really hard running as fast as we possibly can with a bunch of incredible people that have been extremely motivated over a long period of time. And we still have a lot of running left to do. So we think that our job is to keep getting better. We think that we should expect people to notice the success that we're having, and we also think that the stuff that we've built in the past that now sits behind us is a pretty big moat. And then there's a lot of stuff left that we're going to build in the future that once we climb that hill, that will turn into a moat as well. So I think we just have to keep running our plays, I think.

Our next question will come from Naved Khan with Truist Securities.

This is Robert on for Naved. I have one for Mark and one for Ernie. So Mark, I think in your prepared remarks, you said you guys expect operational constraints to improve in the fourth quarter. Is there any color you can give on how we should think about unit growth on a sequential basis?

And then actually, either can answer this question. But Mark, you did just touch on it on the wholesale side. I was going to ask if you're seeing increased throughput of vehicles into the wholesale channel due to some of the operational constraints. But also, are you guys seeing higher demand in the wholesale channel because there's demand to stockpile vehicles amid this inventory supply constraint in the market right now?

Sure. So yes, I mean, on the first point, yes, we expect to increase our operational capacity in Q4 with an eye toward 2022. So it's always a goal, and Q4 is always a very significant investment period for us as we work to ramp operational capacity in preparation for the first half of the following year. And that's exactly our plan this year.

On the second question, I think the -- so our wholesale volume has been very strong, and our wholesale GPUs have been strong as well. I do think that's partly due to a -- well, particularly the GPUs, are partly due to a strong wholesale market. There's been a lot of commentary out there, which I'm sure you're aware of as well, about -- that we've seen a lot of appreciation in the wholesale market this year. And so some of the dynamics that you outlined are certainly playing out. I think we've also made really great fundamental improvements in the way that we are able to buy cars from customers, both from an awareness perspective as well as from a process and technology perspective. So that's certainly driving some of our volume as well.

Our next question will come from Seth Basham with Wedbush Securities.

My question is around the metering again. Have you eased on the metering in the fourth quarter yet?

We have not materially eased on the metering of -- there's a couple of steps that we took that we probably have pulled back on a bit, but not materially. And again, I want to restate, I do think the moves that we made to meter sales were real and they were an impact. But the bigger impact is just the constraints that already existed in the system before. So hopefully, that's at least directionally helpful.

Okay. No, that's helpful. And we found a lot of the metering, at least as it relates to the number of units shown to customers on the West Coast markets, to be more severe than other areas of the country. What is the reason for that? Is that because of the lack of proximity to IRCs or something else?

So I'm going to try to answer that question more generally. What I would just say is I do think as it relates to the particular move of impacting the number of cars that show up in a search result, we do have the ability to kind of provide more relief to different groups based on the cars that we choose to suppress. And so if we have really busy logistics legs, then maybe cars that would need to traverse those logistics legs to get to a certain market, they may not be displayed in that market when we're trying to handle a situation like this. If we have certain classifications of cars that require more work from certain operational groups inside of customer care, then we may be more inclined to suppress that inventory in more locations so that less of that type of work needs to be done, if that particular customer care group is behind.

And so we really can operate that, I don't want to say perfectly surgically, but fairly surgically. And so it's been -- it's a high-quality tool for us to use in a situation like this. And I do think the degree to which we use it does vary by location and by classification of inventory type.

Our next question will come from John Colantuoni with Jefferies.

I have 2. Just wanted to start with the consumer purchases. Maybe you could talk about how much of the increase in consumer purchases is just out of necessity because of a tight wholesale market versus your view that it's just simply a more profitable and efficient channel for inventory? And related to that, once new car manufacturing gets back on track, do you envision pulling back on customer purchases perhaps to a point where it's back towards the levels that you were targeting at your Investor Day as a percentage of total sales? And I have a follow-up.

Sure. Well, you're not adhering to the one plus a follow-up rule then, but we'll do our best to answer. So I think first and foremost, on buying cars from customers, I think that's -- necessity would not be the way that I would describe it. I think the way that I would describe it is it's better. It allows us to provide a high-quality experience to a customer that we're buying a car from, and then it gets us access to a higher-quality pool of inventory that is, on average, more profitable. And so we want to do as much of that as we possibly can. We want to build the business capacity to be able to handle as much of that as we possibly can.

Now I do think that we're in a very unique market today where acquisition channel is very, very important. And normally, across time, if we go back to this kind of 20-year view that we talked about before, if you look at different acquisition channels, you tend to see margins that are available that are fairly stable. There can be kind of idiosyncratic differences between the channels. But generally speaking, they're fairly stable. I think today, the best source is probably customer lease returns. There's -- Cox Automotive keeps an index on the amount of equity that lease returns have. And recently, that number has eclipsed $8,000, which is an enormous number, right? That number is normally closer to 0. So that means the customer is returning a car and kind of dropping off the keys and walking away from a lease. That car has $8,000 of positive equity in it. So that's an enormously advantaged channel in this environment.

On the other side, you have auction, which has been heavily constrained in this environment. There are many dealers that are constrained in inventory. And as a result, many of the off-lease cars for the reason that I suggested before and for the reason of dealers being constrained aren't making it all the way to auctions. There's been a fewer repos because there's been really high-quality customer credit performance. So there's less of that showing up in the market. The rental fleets have struggled to get cars, so they have been defleeting. So there's been less of that in the market. And so that market has been very tight. And as a result, the margins that are available in the auction market are very small. And so that means that kind of the channel of buying cars from customers is even better today than normal.

Our expectation, again, just given the kind of very strong and persistent forces that have existed over a very long period of time in this market would be that as all these idiosyncratic, strange things that are going on in the world today start to normalize, we'll probably see these markets move in more normal ways relative to one another kind of to how they have in the past. I don't know that, that changes a ton for us because I think today, we're in the lucky position to be able to acquire many cars from customers and also have access to auctions. I'm not sure it changes a ton for us. But I think that as that happens, I can imagine the best answer is for us shifting to some degree, and we'll react as intelligently as we can in those moments.

And one quick one. It looks like compensation expense per unit sold was up 30% year-on-year. Just talk about if that's a reflection of higher employee costs from the capabilities you're building around customer acquisitions and how we should think about the trajectory of compensation expenses.

Sure. So the largest component of the compensation expense per retail unit that we saw in Q3 relative to last year, it really is just building the team both in advance of 2022 and just making sure that we're prepared for the first half of 2022 and also for the long term. I think we see so many opportunities throughout the business in so many places where we believe we can invest in the team that can help us scale the business and help us take advantage of some of these opportunities. So those are definitely components.

I think buying more cars from customers was a component as well. As I mentioned, there's some compensation expenses there that show up in SG&A. And then the last thing that I would point to is we were operationally constrained. And so when you're investing for next year and you're investing for the long term, but you're operationally constrained and meter sales as a result, that can also lead to a per unit impact. So those are some of the primary points that I would call out.

Our last question will come from Edward Yruma with KeyBanc Capital Markets.

You guys have been really clear about some of the backlogs and reconditioning. I wanted to ask a little bit, though. There's been some negative press around customer service issues. How do you feel about the strength of the organization today as it relates to customer satisfaction? And do you think that, that's something you have to look at metering as well?

Sure. So I think first and foremost, delivering great customer experiences is why we're here, and it's the most important thing that we do. So it's something that's always at the absolute top of our mind in everything we do. I think throughout these constrained periods, we've continued to deliver customer experiences that get rated much higher than anything else in automotive retail and at levels on par with some of the best customer experiences out there.

That said, across time, across the years and certainly recently as well, when we are constrained and to use the language I used earlier, our service levels get impacted, that does lead to the directional impact that you would expect in customer experiences. And so we have seen that. And that also absolutely went into the rationale for choosing to proactively meter sales while we kind of reduce pressure and caught up. And we've seen that have the effect that we would like to see. We've seen a pretty strong move back towards the levels that we more traditionally have been at. We still got a little bit of room to go there, but we've seen a strong move.

So I think throughout, the customer experiences, in general, have been great. But they definitely are impacted by constraints when we have them. And so it's our job to alleviate those constraints and to build the business in a way that puts no impact whatsoever on customers. But we have had some impacts. And of course, we can continue to get better.

This concludes our question-and-answer session. I would like to turn the conference back over to management for any closing remarks.

Perfect. Well, thank you for joining the call, and thanks, everyone, on the Carvana team. I always say the same thing here, and I apologize for not being more creative, but this was another great quarter. We faced a number of challenges that we didn't foresee. And once again, you rose to the challenge and put us in a great spot. And that's why we continue to be in the place that we're in. And that's why we keep marching towards the goals that we mutually share. So thank you so much for everything that you're for doing. We wouldn't be here without you. Talk to you all next quarter. Thanks.

The conference has now concluded. Thank you for attending today's presentation. You may now disconnect.