Good morning, everyone. Thanks for coming. My name is Rajat Gupta, member of the Automotive Equity Research team at JP Morgan. Very pleased to have with us Mark Jenkins, Chief Financial Officer at Carvana, as well as Meg Kehan, Senior Director, Capital Markets and Investor Relations. I believe Mark has a few prepared remarks and a slide deck to go through, after which we'll go into Q&A and also give the audience an opportunity to ask some questions. With that, thank you, Mark.
Thank you very much, Rajat. Happy to be here again. I think we've been at this conference several years running now, so thank you for having us. The presentation I want to walk through today is, it'll start with a little bit of background, which I think most people in the room are probably familiar with, but what the real goal of the presentation will be is to dive down into some of the very significant fundamental efficiencies that we've gained in the business over the last year or so. We'll really focus in on retail and wholesale GPU, where we've made, I think, really significant fundamental gains that showed up in our Q2 results. We haven't spent much time so far talking about quantitatively and talking about in detail, what's really driving all of that progress?
That-that'll be the primary goal of today's presentation. With that, I'll, I'll start to get into it. I think, here's, here's a safe harbor for forward-looking statements, which we always include. The, the, the two slides on background. I think most people probably have a pretty good s-sense of Carvana's recent history. You know, over our first eight years as a company, from 2013 to 2021, we were one of the fastest-growing companies of all time. I think we were one of the 4 fastest companies ever to join the Fortune 500, with only Amazon and Google growing faster. You know, I think we really focused on growth in the business for our first 8 years, and we're very successful.
Growing market share, growing revenue, growing units very, very quickly. 2022 was a bit of a transition year, so we overbuilt the business for 2022. I think the demand environment rapidly changed in the used vehicle industry and for our business, with very high used vehicle prices and rapidly rising interest rates. You know, in overbuilding the business in 2022, we took a very significant step back in our financial results. Which, again, I think everyone is aware, but, you know, GPU fell for the first time, Adjusted EBITDA losses increased significantly, and EBITDA margin fell for the first time. We really switched our focus at that time toward a shift toward focusing on operational efficiencies and profitability.
I think 2023, moving to this year, is showing the fruits of some of that shift in priorities and our really intent focus on making sure that we're building a very efficient, profitable business that can scale profitably and grow profitably when we ultimately decide to return to growth. 2023 has been very successful on that front. You know, we set first quarter records on Total GPU and Adjusted EBITDA margin in Q1. We set all-time company records for Total GPU and Adjusted EBITDA in Q2. We expect to set records on those two metrics, third quarter records, for those two metrics in Q3 as well, which I'll talk a little bit more as we go through this presentation. 2023 is really, you know, I think has been a very successful year so far.
We completed, you know, the first goal in the three-step plan we've laid out. We achieved positive Adjusted EBITDA in Q2. We've moved on to step two of our plan, which is to drive significant Adjusted EBITDA per car sold on a sustained basis. You know, we sometimes have referred to that step two as drive significant positive unit economics, but really what we mean is drive significant Adjusted EBITDA per car sold on a sustained basis. So that's where we are right now. Step three of our plan, of course, is to return to growth when we feel like we are ready.
Okay, just as another, you know, quick background slide, I think I hit most of the key points on this, but 2023 really has been a year of returning to the trend that we have been on over the last decade. For our first eight years in existence, we improved Total GPU for eight consecutive years before 2022. We improved EBITDA margin for eight straight years before 2022. 2023 is back on trend. 2023 will make it nine out of 10 years of improvement on those key profitability metrics. Okay, that's the background. I think everyone's seen our, our recent Q2 results, as well as our Q3 outlook.
I think one of the things that we want to spend a little bit more time talking about here in this presentation is some of the fundamental drivers of the very significant success we've seen in driving profitability in the business, in, you know, in Q2, relative to where we were, certainly last year, but even in 2021, which was our previous high watermark for profitability. We're gonna break this segment into two different sections. First, I'm gonna talk about Retail GPU, then I'll talk about Wholesale GPU. These are some of the areas where we've seen the biggest fundamental gains.
In Retail GPU, I'll break things down into four key areas where we've really made significant fundamental gains: improving cycle times or inventory turn times, lowering costs in the Inspection Centers and logistics network, buying more cars from customers from a vehicle sourcing perspective, and generating additional revenue streams from our fulfillment network. All of those in total sum up to a range of $600-$1,100 of total Retail GPU improvement that we believe is fundamental and sustainable relative to a benchmark of full year 2021. That's sort of the outline for what we'll talk about next. Now I'll dive into some of the details.
The first key driver of our success, or our recent success on Retail GPU, is what I'll call normalizing inventory size, or really it's more about normalizing inventory turn times. The chart here on the left shows our average days to sale metric by quarter from Q1 2021, up to our expectation for Q3 2023. Average days to sale is basically, it's the number of days between when we acquire a car, either from customer or in the wholesale market, and when we sell a car to a customer on a completed retail transaction. You can see, just as a benchmark, in 2021, we were running on average in the low sixties, average across the four quarters.
We really increased this metric, which is unfavorable for Retail GPU, because if you hold a car longer, it has more time to experience depreciation before you sell it, thereby selling at a lower Retail GPU. It stayed really elevated through 2022 when we overbuilt the business for the ultimate sales environment. You can see it sort of in the mid to high 90s there through most of 2022. It was really high in Q1, where we really, you know, put the pedal down on moving aged inventory. You know, the cars we sold in Q1 were on average, 130 days old. That's really old. That's not normal. We've really now started to see the normalization of this average days to sale metric. We improved by 40 days, going from Q1 to Q2.
That is the single biggest driver of the sequential step up that we saw in Retail GPU going from Q1 to Q2, was that big reduction in days. We expect an approximately 20-day further reduction in Q3 relative to Q2, and then we also expect it to drift down into that low to mid-60s range, which is our near-term target, as we look toward quarters beyond Q3. I, I think one of the things, you know, This is one of the useful metrics to help understand some of the fundamental improvements that we're making in Retail GPU. I would say, you know, we've got a 20-day improvement going from Q2 to Q3.
There were some transitory tailwinds that helped Retail GPU in Q2, but a 20-day improvement in average days of sale going to Q3 is obviously a fundamental tailwind that will help Q3 relative to Q2. Even as some transitory tailwinds in Q2 burn off, we've got fundamental gains going to Q3 that will help Retail GPU. The next category of fundamental gain in Retail GPU is just in fundamental cost improvements. This, I would say, this is my favorite slide in the whole presentation, because what this slide captures is a lot of fundamental improvement, not only relative to 2022, when we were very inefficient relative to historical norms, but it also captures very significant fundamental improvement relative to FY 2021, which was our previous high watermark for Retail GPU and Total GPU.
What this chart on the left shows is it's, you know, quarterly from Q1 2021 to our expectation for Q3 2023. It shows the difference between 2021 costs on reconditioning and inbound transport, which are the sort of fundamental items other than vehicle acquisition costs that drive retail cost of sales. It shows the difference between FY 2021 value and what we are on a quarterly basis. There's a few things that I think are really sort of helpful to point out on this chart. One is that our operational focus is paying off. We've made very, very substantial gains since 2022.
You know, in sort of early to mid-2022, our teams really shifted their focus to operational efficiency, and since that time, you can see we've made very large changes to cost per unit in the reconditioning centers and inbound transport network. I'll talk a little bit about what's driving that. The single biggest source of the gains over that time period have been insourcing. Basically, taking, you know, all production in-house. Some of our third-party production was actually at ADESA in 2022. We obviously have internalized that. You know, we have the opportunity to sort of integrate some of our processes as well as internalize any margin that ADESA was earning prior to our acquisition. We've really focused on, you know, also insourcing services at more and more of our IRCs. That's the single biggest driver.
We've also made gains elsewhere. I think, you know, just getting our staffing right, standardizing processes across a nationwide network of Inspection and Reconditioning Centers, making real improvements in developing our proprietary systems that run the IRCs, and determine how the cars go through the process. You know, getting efficiencies and acquiring parts, for example, aided by software. We've made big gains there, and we've also made big gains on inbound transport, basically driven by having more locations where we recondition, which allows us to reduce inbound transit miles, and also just getting better utilization on the inbound transit network. All of those fundamental gains, I think there's a few things that excite us about them.
One, over this period, our vehicle quality metrics have been stable, which means these fundamental improvements are not coming at the cost of reconditioning quality. They're really coming from fundamental gains in the way we're operating the business that we think are sustainable. On that point, we also think these gains are not the type of gains that all of a sudden reverse when you start to return to growth. So, you know, some of, some of the key sources of gains, actually improve when you start to grow, including logistics network utilization and inbound transport miles, right? As you start to grow, you're able to increase reconditioning capacity where you're acquiring cars, thereby, you know, lowering the inbound transport distance for certain cars that you acquire.
The others we think are real fundamental, sustainable gains that don't require reversal when we ultimately return to growth. Those are some of the qualitative points. You know, we feel really good about this progress. I think it was an amazing success for the teams involved in it, and I think bodes really well for our outlook on Retail GPU and our outlook, when it's time to start scaling volume again. Diving into a couple more of the numbers. You'll notice in this slide, one of the drivers of the sequential improvement in Retail GPU going from Q1 to Q2 was, some of the cost efficiencies flowing through.
In this chart, you can see the step down from $180 relative to FY 2021 to -$180 relative to FY 2021's, about a $360 sequential improvement in costs per unit sold. We do expect further improvements in cost per unit sold on cars sold in Q3. You can see we expect to go from, you know, -$180 relative to FY 2021, down to a midpoint of -$350 relative to FY 2021. We do expect further sequential gains on this metric in Q3 relative to Q2. As I said, with days of sale, you know, this is a real fundamental gain that will help Retail GPU in Q3.
There were some transitory benefits that helped it in Q2, we, we-- you can say this is real fundamental gains that are, are going to help sustain a strong Retail GPU as we look forward in the coming quarters. Last, I, I just hit on the numbers. I think a really helpful thing to keep in mind, that in Q3, we do expect to be $300-$400 per car sold better on reconditioning and inbound transport costs than we were in FY 2021. That's real, fundamental, sustainable gain. Moving on to the third of our four drivers of Retail GPU. Over the years, we've had great success with vehicle sourcing. We're -- we source a lot of our vehicles directly from customers.
We have an amazing customer offering from a customer experience perspective, where people can get a real value in minutes on their phone or desktop, and we'll come to the car, pick up their car, and hand them a check. That's an incredible customer experience. It has had a lot of uptake with customers, and it's led to steady gains in customer sourcing over the years. Our customer sourcing ratio is basically the percent of retail units sold that were sourced from customers versus B2B in the wholesale market, is about 83% so far this year. That's up 10 points from 2021. We think This kinda, you know, mid-80s, low to mid-80s range is around the upper end of our target range.
You know, we don't see ourselves pushing to the point where we're, you know, all cars sourced from customers. We do think it's useful to source cars elsewhere in the wholesale market. We, we don't expect this number to, to drift up much from here. I think our future gains in this area will be focused on customer experience and just improving unit economics by making sure we're buying the right cars. In total, relative to FY 2021, which is the benchmark for this discussion, we view this as approximately a $100-$200 fundamental benefit. The last source of fundamental gains in Retail GPU that we see today relative to where we were in FY 2021 is on driving additional revenue streams through our logistics network.
One of the big advantages we believe of our model is the ability to make a very large selection of cars available to customers conveniently and cost effectively. The chart on the left actually highlights some, you know, metrics around that and the way that we think about our selection advantage. Again, this is enabled by the fact that we've built, you know, this consumer-facing first-party logistics network and invested very significantly in the network itself, as well as the technology that backs the network. On the left-hand slide, you know, we picked our oldest market of Atlanta and just looked at the local dealer inventory in that market as proxied by the listings that show up on CarGurus platform. It's about 50,000 cars that show up in the Atlanta market.
Carvana today makes available about, just over 25,000 cars in that market. The cars that can be delivered to customers' door in Atlanta are 50% as many cars as the entire Atlanta local dealer market, as proxied by CarGurus, CarGurus listings. That's a big, big selection. Now, if you think about, you know, if we're 50%, if you add us plus the market, you know, we're about third, a third of the available cars, at least, you know, across those two channels. We, in our view, that points to, over time, the opportunity to take very large market share as, customer behavior normalizes around online car buying. At any rate, I think the main point today is, you know, we have a very big selection that we can offer to customers in Atlanta.
We also have a very big free shipping selection. We have started to charge shipping fees on longer distance shipments. That's generated about $300 of incremental revenue relative to FY 2021. We're able to do that because we've built this first-party network. But importantly, we still have a very large selection of free shipping cars available to customers in our markets. Using Atlanta as an example, we have about 3 times as many cars as the next largest alternative dealer. You know, we do have the largest selection of free shipping cars available in a market like Atlanta, which we think is really great from a longer-term perspective. All right, main... I think we've hit the main takeaways here. The next slide is just gonna sum up.
It's, it's just gonna sum up what we talked about on the previous slides, and, and this is just building up to that $600-$1,100 of fundamental gains in Retail GPU that I talked about in the intro to this section. You know, see here, we've got $300-$500 of lower costs. The $500 is a little bit higher than the $400 that we had at the top end of our Q3 range, 'cause we do see further gains that we have the ability to make by the end of the year. Expanded customer sourcing, and additional revenue streams come from the previous slide, just with a range around them. You know, FY 2021 Non-GAAP Retail GPU is about $1,700.
Off of that baseline, you know, we feel like we've made in the business, you know, really significant fundamental gains that supports much stronger Retail GPU in a way that is sustainable and can, you know, can persist. I think we've done it for one quarter. I, I think we recognize that it's gonna take multiple quarters of demonstrating this new Retail GPU model, you know, for it to be fully appreciated. We do feel like we've made really significant fundamental gains that allow sort of a new level, a new persistent level of Retail GPU. Okay, that wraps up my comments on Retail GPU. We've talked about the drivers that I just walked through qualitatively in our materials, including our shareholder letters.
We wanted to provide some additional quantitative data to help make things a little bit more tangible in terms of the, the gains that we've made. Okay. Now I'm gonna move on to Wholesale GPU, which is another area where we've made very significant fundamental gains in the business since 2021. I think those fundamental gains are led by the acquisition of ADESA, which does two major things for us. First, it generates an additional stream of gross profit and EBITDA just from the ADESA business of third parties exchanging cars through the ADESA platform.
In addition, the acquisition of the ADESA business and the 56 physical auction locations that came with ADESA enable us to have a infrastructural base for expanding our business of buying cars from customers and selling them in the wholesale market to dealers that want to, you know, buy, recondition and retail those cars. That's what I'll talk about in the next few slides. You know, just, you know, kind of making a couple of those points quantitative. On the ADESA business itself, I think 2023 has been a really good year from a, you know, an improved growth and improved profitability perspective. 2022, you know, was a down year for the industry as a whole from a volume perspective, is also a down year for ADESA.
I'll give a little bit more data on that in two slides. you know, we're growing volume at ADESA this year. I think total growth so far this year is, you know, about 10% year-over-year. That reflects a market share gain based on industry data sources. ADESA is outgrowing the broader auction industry based on the industry data sources that we use. Then, I think ADESA also is now generating positive Adjusted EBITDA, with an opportunity to generate meaningfully more positive Adjusted EBITDA over time. I'll talk about that in a couple slides. In terms of the impact on Carvana's total business, since 2021, I think that can be captured by the chart on the right.
Wholesale GPU, combining the third-party marketplace platform, as well as our platform of buying cars from customers and wholesaling them, average about $450 per retail unit sold in 2021. That stepped up to just over 1,200 in the first two quarters this year. We expect that to step down in Q3, primarily because of the very significant wholesale market depreciation. You know, we've been seeing, but still, at Q3 and certainly at year-to-date levels, a big step up in this area relative to FY 2021, that we think is sustainable, and even more than sustainable, we think is growable over time. That's what I'll talk about on the next couple of slides. The wholesale platform is just an...
It's an area of the business where we do see a significant growth opportunity looking forward. I'll again break it down into those two categories. In terms of the next two slides, these categories will be, one, cars that we acquire from customers and sell to dealers in the wholesale market. We've made a lot of progress over the years. Over the last three years, we've grown that volume of buying cars from customers and wholesaling them in the wholesale market by 60% CAGR, up to 170,000 annualized units. However, that leaves a huge gap between us and the largest player in this market, which does, you know, in the trailing 12 months, about 560,000 units, growing at 11% CAGR.
This is an area where we see meaningful opportunity to expand our business now that we've taken on ADESA. We have the locations to be able to intake the cars, inspect them, store them, get them out to dealers that may want to buy, inventory that Carvana can source from its customers, given its brand, its last-mile delivery network, and its infrastructure for being able to, to stage and wholesale these cars. We see this as a big opportunity. It's an opportunity that we plan to go after. We put some math up here, you know, at a target $800-$1,000 per wholesale unit sold, matching the largest player, which is a stretch goal, but one that we plan to go after, generates $300 million-plus of annualized EBITDA. All right.
The other part of the wholesale platform is ADESA. I think there's, you know, a few things that, you know, I think we're excited about here. One, this year, ADESA has started to generate positive EBITDA. It's $23 million of positive EBITDA year to date. That's not including any of the synergies with Carvana. For example, the cost synergy that I described earlier with, you know, internalizing reconditioning margin at ADESA per cars that they re-recondition on our behalf. That's a cost synergy that impacts Retail GPU, but wouldn't show up in ADESA EBITDA. The wholesale transport, inbound transport benefits that we talked about in one of our previous shareholder letters, another cost synergy that's significant in our integration with ADESA, but wouldn't show up here in ADESA EBITDA either.
This is a number that is just the ADESA third-party marketplace business without any revenue, gross profit, or synergies with Carvana. And so I, I think, you know, that's a positive step from 2022 in the first place. Moreover, I think, you know, one of the things that, that we're optimistic about is 2022, from a volume perspective in the industry, was a really depressed year. The chart on the left here shows, from a industry data source, the, you know, national auction volume annually. And you can see the auction industry has gone through ebbs and flows, stepped down meaningfully in the financial crisis and then rebounded, stepped down significantly post-pandemic, but has started to rebound in 2023.
Just, the industry sort of growing cyclically, is a meaningful contributor to ADESA EBITDA over time. Of course, you know, ADESA just growing with the industry, doesn't count our opportunities for operational efficiencies at ADESA or for taking market share. So the-- I think the, the opportunity for us at ADESA is even bigger than just an industry rebound. We think we have the opportunity to take share, and indeed, we are taking share so far in 2023. So I, I think those are the two, I think main points that I wanted to share on Wholesale GPU.
The main takeaway there is our Wholesale GPU, driven largely by the acquisition of ADESA and, and what it does both in its own business and for our wholesale vehicle business, has, you know, caused a meaningful step-up in GPU relative to 2021. Moreover, we see this as a growth opportunity with an opportunity for additional gross profit and EBITDA contributions from these two businesses. That wraps up my comments on Wholesale GPU. I have one more slide, which is an update on Q3. In, you know, our recent shareholder letter, we provided an outlook of over 5,000 GPU and positive Adjusted EBITDA. We said there was some upside to that outlook, but given how early in the quarter it was, we went with 5,000 and positive Adjusted EBITDA.
We are updating that outlook today. We're updating the outlook on GPU and Adjusted EBITDA. On GPU, we're updating to over $5,500. On Adjusted EBITDA, we're updating to over $75 million of Adjusted EBITDA in Q3. I think for putting that into a little bit of perspective, in Q2, we had some one-time items that impacted our Adjusted EBITDA. We had some benefits from previous valuation adjustments on our inventory that positively impacted it. We also sold more loans than we originated, which positively impacted in Q2. In Q3, our outlook assumes no significant impact from inventory valuation adjustments. We think we're back into a normalized world there, unless something really changes quickly and dramatically in the industry.
The outlook does include some loan sales in excess of the loans that we originate. To be specific about that, it assumes $300 million-$500 million of incremental loan sales above what we originate in Q3. Just to put a little bit more context about the outlook. With that, that wraps up my presentation, and we can hand it over for Q&A.
Great. Thanks, Mark, for the detailed presentation and, and all the color. Very helpful. I just wanted to follow up on, you know, the third quarter guidance, you know, the 5,500+ GPU. If you look at the second quarter, it was $7,900 of one-time benefits. It looks like you're still gonna have a little more backlog clearance on the finance GPU. You know, my math would suggest, like, there's still, like, $300-$400 of benefit from that. You mentioned the $250 of Retail GPU going away, but there's, you know, the day sale benefit, you know, that would equate to somewhere like $300, you know, if it's 20 days better. There's a recon cost benefit around $150.
The math would imply something close to, like, $6,500, you know, despite, you know, all the one-timers. What are we missing between the $5,500 and the $6,000, or just leaving some more room for seasonality or, or anything else that we might be missing?
Sure, yeah. I, I think the one thing, the one thing I, I alluded to a couple times and I think is worth calling out. We don't go through exactly that process. If we just start with Q2 GPU and add everything I walked through in this deck, we do think Q2 had some tailwinds. One, I think we think Q2 is typically a strong quarter for Retail GPU. You know, just even a sort of a normal, seasonal tailwind would, would create some tailwind in Q2, that might not be as strong in, in Q3. Then we also think Q2 benefited a little bit from, -
some transitory market dynamics, in particular, just, you know, wide spreads, and low wide spreads between wholesale and retail prices and low retail lower than normalized retail depreciation in the quarter. I think that's an important thing to keep in mind. You just, just don't take our, our Q2 GPU and add everything I just walked through. I think more of what we're trying to get across is, yes, Q2 benefited from some, you know, seasonal and market dynamic tailwinds, but there's still a lot more in the tank, both on fundamental improvements in days of sale, because Q2 cars were 91 average days of sale, which is well above normalized and on costs, where, you know, we have very clear visibility into another significant step down in cost in Q3 relative to to Q2.
That's the way I would, I would talk about all those things balancing out.
Got it. Got it. That's helpful. Maybe I'll ask one more question and give the audience an opportunity to ask another one. You know, you've completed step 1 of the plan. You know, you mentioned you're ready to move to step 2. If you look at the second quarter results, you know, even taking out, you know, some of the non-recurring benefits and seasonality, it's still like a pretty healthy EBITDA per unit level, you know, like, call it like $1,000, like, give or take. Is that not significant enough for you to move to step 3, or would you like to take this higher before you start to return to growth? Because step 2, you mentioned significant unit economics before we return to step 3, which is growth. Aren't we already there with, like, $1,000 EBITDA per unit?
what's, what's kind of like the plan there?
I want to stay away from setting a specific dollar per unit target for when we return to growth. We do want to maintain some flexibility there as opposed to having, you know, a very specific number. I think the way that we're thinking about this is in step two, we still see significant opportunities to improve Adjusted EBITDA per retail unit sold on a, in a stable volume environment before returning to growth. The, you know, the reasons we see that are we've made an awful lot of gains, in the last year. I think you, you know, you've seen those in our results.
I think from Q1 2022 to last quarter, you know, we improved by, you know, on the order of $500 million of quarterly EBITDA, closer to $400 if you get rid of some of the one-time benefits that we had in Q2. We're really making a lot of gains, and our teams are really focused on driving profitability, and it's going well. We want to continue it. Just from a prioritization perspective and a company focus perspective, we see more opportunity ahead, even on stable volumes. We want to allow the teams to continue to tackle those opportunities. I, I think that's the leading motivator for us staying in Q2. I would also add, when we return to Q3, our goal is to have that be profitable growth.
We get the question sometimes: "Hey, when you start to grow again, are you going to see significant step back on all of these, you know, efficiency improvements that you made?" I think the, the simple answer to that is, that is absolutely not the way we think about it. Our goal is to have growth be profitable growth, and one of the sources for that profitable growth, and one of the reasons we think we can achieve profitable growth is we are operating today with a very significant base of fixed category costs. I think about things like Corporate expenses, Technology expenses, and Facilities expenses. We think we have all...
That's a significant part of our SG&A today, and we think we have the ability to meaningfully grow units into that corporate technology and facilities cost base, which makes growth a meaningful leverer of SG&A when we finally get to step three and return to growth. That's a little bit of extra color, you know, to talk about why we value stage two and why we want to stay in stage two. We're also optimistic about phase three and what that can mean for SG&A leverage, because we do have an elevated and significant corp tech facilities, you know, component of SG&A expenses today.
Got it. That's, that's helpful. Maybe one quick question from the audience, as we're out of time here. Anyone? No, I guess, you know, we'll end it there. Thanks, Mark, for doing this, and thanks, everyone, for joining.
Thank you, everyone.