Good day, and welcome to the Carvana fourth quarter 2019 earnings conference call. All participants will be in a listen-only mode. Should you need assistance, please signal a conference specialist by pressing the star key followed by zero. After today's presentation, there will be an opportunity to ask questions. To ask a question, you may press star, then one on your touchtone phone. To withdraw your question, please press star, then two. Please note, this event is being recorded. I would now like to turn the conference over to Mr. Michael Levin, Vice President of Investor Relations. Please go ahead.
Thank you, Sean. Good afternoon, ladies and gentlemen. Thank you for joining us on Carvana's fourth quarter and full year 2019 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 fourth 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.
The 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. 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, including, but not limited to, ex-Gift measures that exclude the impact of the 100,000 milestone gift to our employees. 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.
Please note that all gross profit, SG&A, and EBITDA metrics mentioned by us on the call today are on an ex-Gift basis. Now, with that said, I'd like to turn the call over to Ernie Garcia. Ernie?
Thanks, Mike, thanks for that extremely lively reading of the risk factors. We appreciate it. Thanks to everyone else for joining the call. 2019 was another incredible year for Carvana, with broad-based gains in all key operational and financial metrics. In the shareholder letter, we once again published our key metrics and our updated cohort graphs. Please take a moment to look over these charts. The story they tell is compelling. It's a story of extreme growth consistently sustained over a long period of time. The magnitude and consistency of these results are only possible because we have built a better business model that delivers the best customer experiences available when buying or selling a car. 2019 was our sixth straight year of triple-digit revenue growth, with over 177,000 cars delivered to our customer, just shy of $4 billion in revenue.
During the year, we delivered more cars to our customers than we had in our entire prior history combined. It isn't just the percentage growth that is noteworthy. At this scale, that growth rate means we sold an incremental 83,000 cars this year. That is the most organic growth of any automotive retailer ever in the U.S. This year, we expect another year of market-leading growth. The powerful unit economics inherent in our model continue to show up consistently. In 2019, we increased GPU by $750. Even more remarkably, it was our sixth straight year of over $400 of improvement in that metric. This year, we expect to notch our seventh. We also saw another year of very strong leverage, with EBITDA margin improving over 4%, lighting the way to achieving our long-term financial model.
We achieved this while simultaneously managing the cost to support 89% unit growth in our retail business and 230% growth in our business of buying cars from customers. Taking a longer view, we have now levered nearly 20 points in the last four years. This year, we expect to continue that march. In November of 2018, we hosted our first Analyst Day. On that day, we covered many topics, but one area of discussion was our plans for buying cars from our customers. We knew it was an important area of focus that would dramatically improve our access to inventory and cut costs from the system, bringing our long-term unit and margin goals into much clearer sight. We knew that we had a very strong foundation to build on top of.
All the assets we're building to deliver cars to customers are the same assets necessary to perform the reverse transaction in a whole new way, which we believe customers would love as much as they love buying cars from us. What we didn't know was just how many cars we could buy from our customers or how long it would take us to make significant progress. At the time, only 16% of the cars we sold had been purchased from other customers. Looking at the numbers of the best-in-class companies in our space and taking into account the quality of our offering and our belief in our own ability to execute, we put out an ambitious long-term goal of sourcing 38%-52% of the cars we sell from our customers.
In the fourth quarter, just four quarters after putting those goals out, we sourced 43% of the cars we sold from our customers. This is potentially the most remarkable progress we've made in any area of the business in any year. It speaks to the fundamental power of the brand, the technology, and the infrastructure we are building. While there are short-term costs associated with expanding that business so quickly, it sets us up very well to achieve our long-term goals. We launched Carvana just seven years ago. In that short time, we've made a lot of progress, and we've made it while simultaneously making the investments necessary to support our current growth and while building the foundations to enable many years of significant growth in the future.
This progress is the result of our focus on our customers, the long-term perspective we bring to everything we do, and the incredible people that pour so much of their energy into helping us build better and constantly improving solutions. Thank you to all those people. Thanks for believing in what we are building together. Thanks for working so hard to deliver better experiences to our customers. Thank you for caring. The machine we're building is becoming easier for others to see from a distance.
Today, we have about 35,000 cars for our customers to choose from. This is up from about 7,500 when we went public three years ago. By the end of the year, we expect to be serving about 250 markets and three-quarters of the U.S. population with as soon as next-day delivery. This is up from 21 markets and just 1/5 of the population in Q4 2016.
We've built the technology and supporting nationwide supply chain infrastructure to allow customers everywhere to buy or sell a car on their time with the best selection and the simplest and most transparent transaction available anywhere. Because we've invested in the fundamental economic advantages that result from a first-party supply chain and a vertically integrated e-commerce transaction, we are also able to take those benefits and redouble them by investing in a culture of true customer centricity. This culture is what powers our offering to be the best and the most fun experience available when buying or selling a car. We are well on our way to becoming the largest and most profitable automotive retailer, to selling 2 million+ cars per year, and to achieving our mission of changing the way people buy cars.
While we do not believe we are a great company yet, our standards are far too high, it is far too early, and we are not the ultimate arbiters of that. We do believe we're on the path to becoming a great company, and while there will undoubtedly be bumps along the way and many risks inevitably remain, we are steadfast in the convictions that have powered us to this point. The chance to build something great is a chance few companies get, and we will scrap and fight every day to stay on this path. We have the right plan and the right people. We're ambitious, we're motivated, we're energized, and we're still just getting started. Mark?
Thanks, Ernie. Thank you all for joining us today. We're pleased to report another quarter of strong growth in both retail unit sales and revenue. Retail units sold totaled 50,370 in Q4, an increase of 82%. Total revenue was $1.1 billion in Q4, an increase of 89%. In 2019, we completed our sixth consecutive year of triple-digit revenue growth. Revenue totaled $3.94 billion, an increase of 101%, and retail units sold totaled 177,549, an increase of 89%. Our exceptional growth in 2019 was driven by rapid growth in our market cohorts. Our existing cohorts grew by 84%. Our four oldest cohorts, each more than three years old, grew by 50%, and our oldest cohort of Atlanta grew by 18%. Our cohort growth was broad-based, with many markets crossing key milestones.
As of Q4, 90% of our markets were ramping faster than Atlanta at the same age, and 23, seven, and two markets surpassed 1%, 1.5%, and 2% market penetration respectively. These trends are a powerful demonstration of the positive feedback in our model with inventory selection, an expanded IRC footprint, national brand awareness, and great customer experiences driving gains nationwide. Total GPU increased by $750 in 2019 to $2,883. Our GPU gains were also broad-based, reflecting increases of $325 in retail, $308 in finance, $81 in ancillary products, and $37 in wholesale. EBITDA margin was -5.8% in 2019, an improvement of 4.1%, reflecting gains in both GPU and SG&A leverage. 2019 was an exceptional year for buying cars from customers. In 2019, we acquired 104,000 vehicles from customers, an increase of 231%.
When combining total retail units sold with standalone vehicle purchases, we transacted with over 256,000 customers, an increase of 138%. The exponential growth we saw in buying cars from customers in 2019 brought many adjustments throughout the year. In Q3, the outsized growth in buying cars from customers led to pinch points in our last mile delivery network, impacting retail unit volume. In Q4, the rapid increase in inventory diversity brought on by buying cars from customers necessitated learning and adjustments to our approaches to bidding, pricing, marketing, staffing, and reconditioning. While these adjustments led to some transitional costs, these costs were small compared to the enormous magnitude of the opportunity in front of us to change the way people sell cars. Buying cars from customers also brought the significant benefit of increasing the selection of cars available to customers on our website.
From Q2 to Q4, our inventory diversity increased by nearly 50%, driven by customer-sourced cars. The net impact of this increase in diversity was a reduction in the average selling price of our vehicles, which we expect to continue and have incorporated into our revenue guidance for 2020. In 2019, we opened 61 new markets, bringing our year-end total to 146. Far this year, we have opened an additional 15 new markets and one more vending machine to reach a total of 161 markets and 24 vending machines. With these new market openings, we now serve 69% of the U.S. population, up from 59% at the end of 2018. We will continue to expand in 2020 and expect to end the year serving 73%-75% of the U.S. population, on our way toward our long-term goal of 95%.
In 2019, we also made significant progress scaling our vehicle production capacity. We added our sixth and seventh Inspection and Reconditioning Centers near Cleveland and Nashville, and began construction on our eighth Inspection and Reconditioning Center near Charlotte, North Carolina, which opened this week. Our goal is to end 2020 with more than 500,000 units of annual production capacity at full utilization. We are squarely focused on achieving our long-term goal of 2 million+ units, and we plan to continue to maintain a healthy pipeline of future IRCs to support our growth. 2019 was an outstanding year for our finance platform. We introduced our first auto loan securitization in the first quarter and successfully executed three more securitizations, further expanding our investor base and recognizing additional gains. For the full year 2019, we increased finance GPU by $308 to $962, compared to $654 in 2018.
We are excited about our progress and expect to make continued gains in our finance program over time. Beginning in 2020, we plan to transition to a two-shelf securitization program from a single shelf in 2019. Our transition to two shelves is a valuable step forward for our securitization program that we believe will unlock an expanded investor base, greater liquidity, more efficient capital structures, and lower cost of funds over time. We ended the year with $460 million in committed liquidity resources and held an additional $67 million of real estate and securities on our balance sheet.
As part of the transition to a two-shelf securitization program, we also held an incremental $110 million of loans at the end of the year that we plan to sell in Q1, and we closed two new revolving facilities with capacity of up to $1 billion to finance loans prior to future securitizations. As we look forward to 2020, we expect another year of significant growth in retail units and revenue, increased GPU, and improved EBITDA margin. Our outlook for retail units sold is 255,000-265,000 units for the year, an increase from 177,549 in 2019. Our outlook for total revenue is $5.6 billion-$5.8 billion, an increase from $3.94 billion in 2019. We expect GPU to increase to 3,200-3,400, reflecting gains across all parts of the transaction, and EBITDA margin to improve to -1.5% to -3.5%, reflecting gains in GPU and SG&A leverage.
Since becoming a public company nearly three years ago, we have made tremendous progress across all aspects of the business. Since 2016, we have grown revenue by 10x , increased gross profit per car sold by more than $1,800, reduced SG&A expense per car sold by more than $1,400, all while providing exceptional customer experiences and building a large and rapidly growing offering of buying cars from customers. We have made significant progress toward achieving our financial goals and are excited about 2020. Thank you for your time. We'll now take questions.
We will now begin the question and answer session. To ask a question, you may press star then one on your touchtone phone. If you are using a speakerphone, please pick up your handset before pressing the keys. To withdraw your question, please press star then two. In the interest of time and to make sure everyone's question is asked, please refrain from asking more than one question and one follow-up. Our first question will come from Zach Fadem with Wells Fargo. Please go ahead.
Hey, guys. Curious if you could walk us through some of the broader assumptions in the unit guide, and then maybe explain in a little bit more detail why you expect unit growth to outpace total sales growth in 2020, and what the implications are for wholesale and other growth.
Sure. I think we're obviously very pleased with the growth that we're seeing across the business today. We put out our cohort charts again this year, which we hope will be helpful as you try to decompose where growth is coming from in growing markets and also growth coming from the markets themselves growing. I think that's what's driving our assumptions. We feel really, really good about those being well-founded assumptions that we feel very comfortable with. Then as it relates to the revenue moving a little bit relative to last year, I think we're trying to leave room in ASP. As we buy more and more cars from customers, we want to really understand exactly how that's going to migrate. We've definitely seen a significant increase in the diversity of cars that we're able to buy from customers.
We grew that diversity by about 50% from Q2 to Q4. Generally speaking, that's kind of a widening of inventory. As that continues to evolve, we want to leave a little bit of room there. Since the major driver of our economic model is units anyway, and we're guiding for GPU, we thought that was a prudent way to guide.
Okay, got it. On the shift to a two-shelf securitization model from a one-shelf, I believe you said in the shareholder letter that this had an impact to Q4 units and GPU. Curious if you could explain that in a little more detail, whether there's going to be an impact to 2020 units and GPU as a result?
Sure. I would clarify one point there. I think it's revenue and GPU, we believe, was impacted in Q4 by holding these loans over as we transition toward our two-shelf securitization next year. Basically where that impact comes from is had we sold the loans, we believe we would have earned a premium that would have impacted revenue and GPU. Instead, we will earn that premium in 2020.
I think that's the primary impact on revenue and GPU from that transition. That would not impact retail units sold. Got it. Appreciate the time, guys. Thank you.
Our next question will come from Ron Josey with JMP Securities. Please go ahead.
Hi, this is Andrew Boone on for Ron. Thanks for taking our questions. You guys now have 35,000 cars on the site, and I think you highlighted 50% greater selection. Can you talk about the benefits of that, as well as just the operational challenges of having more sites at the IRC as well as DSO? Secondarily, where is that coming from? Is there any change in the auction side or is that all just from trade-in? Thank you so much.
Sure. We do have now 35,000 cars on our website. I think something that's pretty unique about our business model that's powered by our supply chain and logistics network that connects all of our markets together, is that all of those cars are available to customers everywhere. That's a major source of positive feedback in the model. As we continue to add markets and grow within markets, we can justify larger inventories, and those larger inventories give customers in every market more and more selection. When we talk about the diversity increase of 50% from Q2 to Q4, that's driven by both the size of the inventory, which has grown. Most importantly, it's also driven by buying many more cars from consumers. Those do come from trade-ins and then also buying cars outright, even when the customer is not buying a car from us simultaneously.
We think that's a tremendous development in the model. If you take a step back and just say, what is the entire market for consumers buying cars? It's all about consumers just trading cars with one another, and there's all these different companies and mechanisms that stand between two consumers that are trading cars, whether it's all the dealers or us playing this role, all the finance companies, the transport companies. There's many companies that sit in between consumers that are trading cars. The more of that system that we can collapse, the more we can cut cost out of the system, the more valuable our platform becomes, the better offering we can give to customers through a simplified, more transparent offering, and the more value we can share with them. We think that that's a huge evolution of the business that we're extremely happy with.
Great. Thank you.
Our next question will come from Rajat Gupta with JPMorgan. Please go ahead.
Hey, good evening. Thanks for taking my questions here. Just wanted to ask my first question on the SG&A leverage. The $60 million in spending that you talked about in the shareholder letter to support this increased retail sourcing, could you give us a sense if majority of that spend is now behind you, or should we expect any more areas of incremental investments in 2020? Just relatedly, looks like the other cost bucket within the SG&A line, it continues to grow. Again, a part of that is likely driven by this retail sourcing initiative, how should we expect that to lever in 2020 and beyond? That'll be all, and I have a follow-up. Thanks.
Let me start with this $60 million incremental SG&A investment in buying cars from customers in 2019. The majority of that was advertising. The remainder was invested in building out our technology, adding to our people for building this business, as well as expenses related to actually executing the transactions. Those expenses, obviously, we think yielded a very good return. The growth in this business has been substantial. We were very excited to hit the middle of our long-term range for customer source cars as percent of total retail units in Q4. We believe those investments are paying off. As we look forward to next year, I think the way to think about it is that the expenses related to buying cars from customers are reflected in our guidance for 2020. We certainly plan to continue advertising the product of buying cars from customers.
We believe that that's a really great thing for us in the long run. We'll continue to invest in the other areas, building technology, making sure that we're appropriately staffed, and the other needs of that business. All that's incorporated in our guidance for 2020.
I'll jump in as well. I think it's also worth, we call it in the shareholder letter, other transitional costs, and I think it's worth outlining those for you as well, because they played a role in the quarter, and they'll likely play a role in 2020 as well. Mark spoke to the direct cost of $60 million, but I do think anytime you have a business that's changing this quickly, there's going to be other expenses that flow through in other ways as well.
Just to quickly try to put a finer point on those, I think there's always opportunity costs. There's management bandwidth, there's prioritization of our tech teams. There's different pinch points that emerged this year that crowded out sales because we were investing so much in this business of buying cars from customers. Then I think another area is just adjustments and learning throughout the supply chain. We've spoken about the diversity increase in our inventory. That definitely drives other expenses that show up indirectly, and you can't point to a line item for exactly what they are, but it means we have to optimize the way that we're bidding on cars that we haven't seen before. It means that we have to run those cars through the inspection center, and maybe we don't have as much experience certifying those cars through the inspection center.
It means we need to price optimize as we sell those cars to customers. There also is a little bit of a marketing mix rebalancing. I do think all of that showed up in Q4 and in Q3. It'll roll into 2020. It's in our guidance. I think to me, something that I think we're pretty proud of, and I want to make sure we point out here, is we also did show reasonable SG&A leverage this year, despite absorbing all of those costs associated with building this business largely from scratch. We bought 100,000 cars from customers this year. That's a lot of incremental transactions that we added. Those transactions don't show up very much. They don't contribute much to revenue.
Through traditional leverage metrics of either dividing by retail units or dividing by revenue, those kinds of investments can largely be hidden a little bit. I think the progress we made this year is something we're extremely pleased with, and we think it's even better than it looks like at first glance.
I'll hit the question on other SG&A. Other SG&A went up by about $166 from Q3 to Q4. Roughly half of that is just seasonal investment. We saw about an $80 increase sequentially going from Q3 to Q4 last year. The rest of that increase, we largely attribute to this rapid transaction growth that we're seeing, driven by buying more cars from customers. I think there's three components in that other SG&A bucket. There's technology, there's corporate overhead expenses, including corporate occupancy, there's transaction expenses. Buying cars from customers impacts all of those, for reasons that we've discussed. We're investing in technology, buying cars from customers. We're expanding our headcount to support that business. There's some additional transaction expenses that go along with supporting that business as well.
I think when you look at our total expenses on a per retail unit basis, as Ernie was laying out, in this phase where we're really building this business up for the first time, that's going to have an impact on per retail unit numbers that we expect to lever over time. Obviously, we're very excited about all the progress in that offering and expect it to deliver significant benefits to the business in the long run.
Got it. Just to follow up on the GPU guidance, the 3,200 to 3,400, pretty substantial 300 to 500 increase year-over-year. Is it possible to disperse out how much of that could come from retail GPU versus finance? Any sense on that? That'll be all. Thanks.
Sure. Similar to past years, we expect gains to come across all parts of the transaction. I think that's been a pattern in our march from 1,023 GPU in 2016, the year before we went public, up to 2,883 today, is that we've been making progress across all parts of the transaction, retail, wholesale, finance, ancillary products. We would expect that to be the case again in 2020.
Got it. Great. Thanks for taking my question. Good luck.
Thank you.
Our next question will come from Seth Basham with Wedbush. Please go ahead.
Thanks a lot, good afternoon. I just want to follow up on the points around buying cars from customers. You guys invested an incremental $60 million in SG&A this year. You bought 104,000 cars versus buying and selling at wholesale 15,000 cars in the prior year. An incremental 89,000 cars that were bought from customers. What is the incremental profit you're earning on those cars? If it's less than probably about $700, I don't see how you can make any ROI on it.
Sure. I think a good place to start at just a very high level is using the numbers that Mark outlined. He said we spent about $60 million in hard costs. I do think as part of building this business, there's definitely indirect costs as well. I think if we're trying to think about how the business rolls forward, probably looking at the hard cost is the smarter cost to zoom in on. If you take that and divide by roughly 100,000 cars that we bought, that implies around $600 of expense per car. I think you can think about it like that. If you look at the full year, our average contribution on a wholesale unit that we bought from a customer and sold at auction, we made around $400, give or take.
I think at a high level, you do all the math there, and you can see where it is right now. Then the question is, where does that go over time? I think we expect a lot of things to move positively there. One, we expect the variable expenses per unit to drop pretty quickly. The largest sum of those hard costs were basically all the investments we made in marketing this offering to consumers. We've seen those expenses lever very quickly in the past on the retail side, and we would expect them to lever very well as well in the future. The remainder of the costs really are pretty small, but there's definitely room for those to lever as well.
On the margin side, if you zoom out and look at the progress we've made on our wholesale units over a multi-year period, you'll see a steady march of progress there. I think that we expect to continue to march that up steadily. Some of the best-in-class operators have that number around about $1,000 as a frame of reference. We obviously expect significant gains in units as well. I think across the board, we view there being a lot of potential there. We think the results that we had this year are extremely exciting, and we're very proud of them. I think the expectation to build that business largely from scratch, to grow it by 3x and to try to have it contribute positively in the very first year, I think is a tough expectation.
We obviously expect it to contribute very positively over time, and we think it's an enormous fundamental benefit to the business that it gives us access to the deepest, most scalable, most diverse source of inventory that exists, those cars that are in consumers' hands. We're really excited about that, and that's roughly how the math works out.
Ernie, when do you expect to turn the corner so that your expenses are less than the incremental profit you're making from buying cars from customers?
I think our results for the sum of the business are taken into account in our guidance. I would point you there, and I realize that's a much more general frame than the question that you're asking. I think it's smart for us to stay at that level of specificity because this is a business that four quarters ago we outlined our long-term goal of being 38%-52%, and then in four quarters we jumped up to 43%. That happened really fast. We gave these other stats that we increased inventory diversity by 50% in the last two quarters. There's just a lot of change happening there. We're seeing cars offered up to us from consumers that we have not seen many copies of in the past, and we're learning how to bid those properly.
If you don't bid those right, you lose cars you should have won, you give up margin in certain places, you give up conversion in others. As they pull through the system, you've got all these different places expenses show up. We're learning very rapidly, I would say, and we expect to continue to make progress. Trying to give you the precise timing of when we're going to see leverage in marketing, that's tough. I think we feel really good about the offering. The results we've seen so far are very strong. We've got a lot of momentum. We feel like we've got a very clear path to our long-term model. I think we're going to point to that, and then as it relates to the business results, we'll just point to the guidance for the entirety of the business.
Fair enough. Very dynamic and exciting. Thank you.
Thank you.
Our next question will come from Chris Bottiglieri with Wolfe Research. Please go ahead.
Hey, guys. Thanks for taking the question. First one is on the, you said, pretty prolific growth of the trade-in ratio. When you look at your retail gross profit per unit in Q4, and whether you frame that versus Q3 or Q4 2018, do you think the increased trade-in ratio had the commensurate impact on the retail GPU that you would've expected based on lower acquisition costs? Or similar to the wholesale GPU, where there are some inefficiencies that maybe mitigated some of the near-term upside that you would've expected?
All right. I can hit that one. We certainly had a very strong quarter for buying cars from customers. I think we were at 1,349 for Q4 2019. That was up more than $400 year-over-year. A big driver of that strength and increase year-over-year was buying cars from customers. We made a nice incremental gross profit on cars bought from customers in Q4. Obviously the ratio of cars bought from customers to total sales has increased significantly, not only year-over-year, but also even quarter-over-quarter, just given the rapid growth in the business. That was certainly a key driver. There's some other dynamics in the retail GPU strength in Q4. For example, Cyber Monday promotion units were a smaller fraction of total Q4 units this year than they were in the past year or last year.
That has a positive impact on retail GPU when comparing the two periods, other things being equal. Definitely some strength in GPU in Q4 2019, and the biggest impact there was buying cars from customers.
Gotcha. That's helpful. Semi-related, you've obviously done a good job of expanding the inventory selection or site via trade-ins. It'd seem like you're testing partner trade-ins on your website and having those fulfilled from other dealers. To the extent you're willing to comment on that, high level, what are the puts and takes and the trade-offs from allowing dealers onto your website? How does it impact the consumer experience? What are the trade-offs in terms of unit economics that, if you do decide to go forward with this? Thank you.
Sure. I'll try to answer that one. What I would say, at any point in time, we've probably got three to five pretty big projects that we're working on. At Carvana, we've got different technology teams working on all kinds of different things, and we have many more than that, other products that we're testing at any point in time. I think generally speaking, our policy to date has been that we're not going to dive into specific products that we're working on or that are going on in the background unless we expect them to impact our results in a material way going forward, or unless they're a big contributor to the results that we've just reported. I think there's always a lot of investment going on.
Many of those things end up working out really well, like FCC and like all the gains we've seen in finance over the last year and a half. Many of those things we roll out and we test and we realize that they don't work great. So I think, as a general policy, I think we're going to stick to what we've stuck to in the past and just know that we're always working really hard to try to make sure that we're coming up with the most innovative and best products we possibly can for our customers to constantly move the experience that they get forward in all the different ways that are important to them. As we have specifics that we feel like warrants being addressed in a venue like this, we will certainly address them then.
Gotcha. That makes sense. Cool. Thank you.
Thank you.
Our next question will come from Rick Nelson with Stephens. Please go ahead.
Thanks. Good evening. Like to dive into the days to sale. Where are you now? What are you targeting?
In Q4, we were 62 average days to sale. That's actually the same as for the full year, where we were also 62 average days to sale. Q4 is up slightly year-over-year. The full year is down by two days year-over-year. I think our thought process on average days to sale right now is that we expect it to stay in a relatively stable and reasonable range. We're very comfortable with the level of inventory relative to sales that we have today. We think we're striking a nice balance between GPU, which is obviously stronger than it's ever been, and making a great selection of cars available to customers on our site, which is really the key trade-off that we're balancing when we're thinking about inventory size. We're very comfortable in the range that we're operating in today.
We do see a significant long-term opportunity to meaningfully bring down average days to sale from current levels. I think our model of having centralized inventory pools that we're able to make available to customers in many, many markets nationwide gives us an opportunity to have very attractive average days to sales in the fullness of time.
Thanks for that. Also, I'm seeing more older, higher mileage cars on your website. Cars 10 years old, 100,000 mi. I'm curious if your return rates on those cars are different. I assume that's coming from your sourcing initiatives.
Let me hit the second part of that first. I do think, as we said before, we've definitely seen a more diverse access to inventory by buying cars from customers, and that is driving a more diverse inventory for our customers to choose from on our website. I think that generally speaking, is widening. We're seeing more expensive cars. We're also seeing older, higher mileage cars, as you brought up. I think overall, year-over-year, I believe our ASP was very similar. That is definitely dynamic as we continue to buy all these cars from customers. We want to keep an open mind toward exactly how that's going to play out. As it relates to return rates and reviews, we've seen very, very consistent reviews and return rates across time, across cars.
I think early on, four or five years ago, a hypothesis would've been that we were going to be selling mostly one and two-year-old cars because those were the easiest cars for consumers to buy sight unseen, and have delivered to their door. I think that's been a pleasant surprise over time, as we found that one, there are consumers that are definitely willing to buy older, higher mileage cars online. That's as fast a growing part of our business as anywhere, maybe a little bit faster even. Two, the transaction technology that we've built, and the processes that we've built, and the supply chain that we've built, work very well for those customers, where we're seeing very high-quality reviews from them as well, and return rates that are in line with the rest of our population. I think overall, we view that as exciting.
We've always thought that our TAM was very, very wide with respect to different types of cars because our offering really is independent of car. We're trying to give customers a broad selection, a great price, a great experience, and that really has nothing to do with the specifics of any given car. We always kind of hoped that would be the case. I think we're continuing to see more and more evidence that points in that direction.
Our next question will come from Michael Montani with Evercore ISI. Please go ahead.
Great. Thanks. One question, one follow-up. The first question was about retention of customers. Now that you look at a market like Atlanta, you're getting kind of five, six, seven years in operation there. Just curious if you could share any percentage of those customers that are actually being retained and returning to buy with you again, just had a follow-up.
Sure. I think we definitely are seeing nice positive signals from our older markets. That said, if you go back to Atlanta five or six years ago, we had very small counts of sales. I think when we look at repeat sales, the data is still thin enough to where I don't think we want to be giving out numerical guideposts just yet. What we're seeing, we really like. We also see a lot of cross-family purchases where one member of the family will buy a car and then other members of the family buy cars. We see a lot of referrals. I think the early evidence for our experience resonating and consumers voting with their feet by coming back again is strong.
I don't know that it's numerically robust enough for us to want to quantify in a way that we feel confident will be consistent across time, given that it's still early in the company's history, and we are only a seven-year-old company.
Okay, makes sense. If I could just follow up on the retail GPU. The increase there was a little bit more than we thought, just based on mix alone. I was going to ask was there any change to the shipping fees or depreciation rates or even reconditioning that would've incrementally supported that we should know? Just on a related note, obviously you mentioned some competitors that have 1,000+ GPU on the wholesale front, but they tend to have their own auctions as well. I'm just wondering if you see potential to get to that level or if you would actually need to open your own auctions to do that.
Sure. Let's take this one at a time. I think Mark kind of walked through some of the contributors to the strong fourth quarter retail GPU. The one other thing that I think I would point to that is useful, especially on a year-over-year basis, is this year our Cyber Monday promotion actually went very well from an operational perspective. We were very prepared going in. I think the team did a great job handling all that extra volume. We were just coming off of some constraints that we had seen in the third quarter, and we knew that we were going to be coming into another year of big growth in 2020. We didn't lean in as much with marketing dollars as we had in the past.
We chose to take a little more careful approach there, just to make sure that we were really positioning ourselves well heading into 2020. As a result, we saw a smaller percentage of our sales in the fourth quarter that got the Cyber Monday promotion discount. I think that was probably also something of a contributor that's worth noting. The other line items that you pointed to, I'm sure moving around a little bit, but I think the two big contributors were mix effects and then the smaller contributor from Cyber Monday. As it relates to auctions, what I would say is we've got a number of great partners we work with today to take our wholesale cars and go sell them, and also to buy many of our retail cars that we're not buying from consumers.
Our instinct will always be to find solutions that work great for us and our partners before going out and tackling another problem and trying to build something from scratch. I think the point is well taken that the best-in-class operators that have a $1,000 margin generally are also monetizing that portion of the transaction, and that's something of which we're well aware.
Our next question will come from Nick Jones with Citi. Please go ahead.
Hi. Thanks for taking the questions. I guess, first, how much higher do you think the percentage of car source can go if you wanted to revisit that target? Do you have a different view after you've seen the demand for the offering? Then, I guess on top of that, when you do source a car from a customer, could you remind us what process do those cars go through? Are you running it through the Carvana checklist? Is there an opportunity to maybe not have to do that quicker? If you are running through that, it seems like, I guess piggybacking off the last questions about auctions, that if you're already indexing the data, that's maybe more valuable through the wholesale channel than some other cars that hit the network.
Sure. Let me take the first one, and then maybe Mark's going to jump in on the second one. That'll be helpful. How high can this go? I think we set our long-term target, I guess now four and a half quarters ago, give or take, at 30%-52%, and that was a thoughtfully arrived at target by looking at what best-in-class operators had accomplished. I think the fact that we got to 43% this quickly is something that we are extremely happy about, and I think, as I said in my prepared remarks, we felt really good about that target early on because we felt like we've built this supply chain that enables a customer experience that we think at least is exactly what customers want. You can go onto our website, we can use all of our different data sources and value that car intelligently.
We can give you a bid, then we can go to your house and we can pick it up and you've got your money and the whole transaction's over. That was something that we were very confident would resonate well. I think that the proof's in the pudding there. It clearly is resonating well, so we're very pleased there. That said, we still definitely have a lot of wood to chop. In an earlier question from Seth, we talked about the expense that we have per unit. That's something that we'll be attacking and addressing over time. We clearly have room in margin as well, getting smarter there and being smarter about bidding those cars and pricing those cars and figuring out exactly which ones we want to buy. There's lots of room for us to grow there.
I think just at this pace, trying to predict the continued speed of progress we're going to make is just very hard. I think that we've definitely benefited to a degree from the brand that we've built up over the last seven years of delivering great experiences to customers, and I think that's probably helped this offering to accelerate so quickly. It's hard to know exactly how quickly it will accelerate from here. I think we're very optimistic about the offering. We're very excited about how fundamentally valuable it is to our entire platform. I think until we really make progress very deep in our long-term ranges across the board, we're going to stick with those ranges as we still feel like we have work to do. We're very excited, as I said.
Yes. Then on your question about process, so after we take a car in from a customer, if it's a wholesale car, we'll typically sell it through one of the major wholesale auctions. If it's a retail car, we'll take it into one of our inspection and reconditioning centers and put it through the normal process that we would use for any car. That includes putting it through 150-point inspection, making sure that the car is reconditioned to our quality standards. We'd photograph the car and post it up on the site, just like a car that we bought through another wholesale channel.
Great. Thank you for taking the questions.
Thank you.
Our next question will come from Brian Nagel with Oppenheimer. Please go ahead.
Hi. Good evening. Thanks for taking my questions.
Of course.
The first question I wanted to ask, I guess bigger picture, but you talk in your letter about the continued rollout of your IRCs, and I think we're working on eight now. As we're looking at your financials and the continued development of financials against the backdrop of some substantial growth, how should we think about the, or factor in just the benefits, so to say, of this IRC rollout? Where is that, so to say, showing up?
I think the number one place that it shows up is in volume. I think as we're marching toward 2 million units, one of the most important, and frankly, most challenging things that we have to do is scale production capacity. Our IRC ramp is focused first and foremost on giving us the capacity to scale our volume on this path toward 2 million units. I think there's potentially some other impacts that you may be driving at. When we launch an IRC initially, it tends to take a little bit of time to get fully staffed up, get up to capacity, as well as there's some learning at the individual site level as people are sort of learning by doing and getting better and more efficient over time.
When we're adding these IRCs, there can be some short-term transitory impact to the reconditioning side of retail GPU. Those impacts have historically been somewhat small, and as I mentioned, transitory.
That's helpful. The second question, just shorter term. Mark, you talked in response to some of the questions about the $60 million investment this year to help support the growth in the effort to buy cars from customers. That's $60 million. If we look at our models, what should we think that number is next year in 2020?
Sure. We haven't given specific guidance on that. I think it is incorporated into our guidance for the full year for the business as a whole. I think, just to break down that $60 million, the majority of it was advertising. We certainly intend to continue advertising this offering of buying cars from customers. 2019 was basically year one of making customers aware of this product. We really started advertising in earnest in the first quarter of this year. Building awareness of this offering is certainly going to be part of our plan in 2020. We'll continue to advertise. We'll continue to invest in some of those other areas as well. Technology, continuing to add people to support this business, and all of that's incorporated into our guidance. Having said all that, we also do over time expect these expenses to lever.
We were in year one, grew incredibly quickly this year, and are very happy with the progress. On the cost side, certainly have expectations of levering over time.
All right. Thank you very much.
Thanks.
Our next question will come from Armintas Sinkevicius with Morgan Stanley. Please go ahead.
Great. Thank you for taking the call. When I look at the cohort curves, Atlanta's tracking at 2.09 versus 1.94 a year ago, so 15 basis points growth year over the last four quarters. The other markets are growing some 30 basis points, so almost double that rate. We've been tracking some of the Google Search trends which suggest Atlanta's been plateauing a bit. Maybe you could talk about that cohort? What have you seen there? Have you seen competitive response after years of taking share, or is it the dynamic around sourcing cars from customers? Anything else that we should be thinking of as we look at the rest of the cohorts and the dynamics there, and as well as how you think about Atlanta going forward?
I would start with this. Atlanta grew by 18% year-over-year when you look at the full year 2019 over full year 2018. That's something that we're very excited about. It grew faster than that when you look at the first three quarters. Then I think in the fourth quarter, due to a couple of reasons that I'll talk about in a moment, I think it slowed down a bit. At that point, you're pretty zoomed in. We're talking about looking at a single market for a couple of months. I think we're trying to put out a lot of data to give you a broader understanding of the gains that we're seeing across the board, because we think it's important to broaden that perspective a little bit.
Something that we do internally all the time is we're obviously looking at individual markets. We've got 160 of them, and we're tracking how those markets are doing month to month all the time. If we see a market bouncing around in a way that starts to look like maybe it's not just noise and there's something to understand there, we dig in, and we try to understand exactly what's going on. We've been doing that for several years, I would say. When we do that exercise generally yields staffing problems or shipping delays or something macroeconomic in that particular city. I can't think of a single time where that has yielded a belief where we think competition is the driver of what's happening in a given city. We really don't think it's that.
We think that Atlanta was probably impacted by the same things that impacted the entire business in Q4. Namely, we leaned into the Cyber Monday promotion less. Then we did have some of these costs associated with the transition to buying so many cars from customers. Those took many forms that I think slowed things down, including crowding out sales, which is a real thing in a market the size of Atlanta. Then circling back to the Cyber Monday promotion, I also think that was actually more pronounced in Atlanta. Something that we've definitely observed over the last several years is older markets and markets where we have more awareness tend to be less responsive to our promotions. I think that's probably because the promotion kind of serves two purposes. One purpose is it offers consumers the opportunity to get a great deal.
Another purpose is that it serves as an opportunity to become aware of our offering at all. I think in markets where we already have high levels of awareness or higher levels of awareness, I think that second value is a little bit more muted. We've seen over the last several years, Atlanta's share of sales in the fourth quarter associated with the Cyber Monday promotion pretty steadily decreasing. I think that was another contributor there as well. Again, overall, I would say that market performed very well in our eyes. It grew at 18%, growing faster than that in Qs one, two, and three. I feel like when we zoom into Q4, we do understand it. We probably won't always understand what's going on at an individual market level. There can be noise.
The growth has been pretty steady, and we're very excited about it.
Okay. When I look at the web traffic in the K, it implies a slowdown in the fourth quarter versus the third quarter. Any insights there? Obviously, still strong traffic, but what should we be thinking about what you're doing to drive traffic and maybe what happened in the fourth versus the third quarter?
Sure. Well, I would start with sales. We saw healthy growth in cars that we sold to customers and obviously very healthy growth in cars that we bought from customers going from the third quarter to the fourth quarter. That happened despite leaning into the Cyber Monday promotion a little bit less. I think when you're seeing variation in traffic that is kind of not correlating super tightly with sales, that can often be just marketing channel mix differences and many other things. Yeah, I wouldn't read too much into that. I think there can be a lot of variability in quality of traffic from different channels, and sometimes you buy a lot of traffic at a low price per visitor, and sometimes you buy high-quality traffic at a higher price per visitor. Just traffic alone is not necessarily the best measure to look at.
That said, traffic has continued to grow very quickly and so we're happy with that as well.
Okay, great. The last one here, just with regards to raising capital, how do you think about the capital structure of the company over the next couple of years?
We continue to be very focused on asset-based financing, which we think is the best way to capitalize the business. I think we've talked about this before, automotive retail utilizes various types of capital to fund its working capital and its other investments. The main two being investment in inventory and investment in real estate. We have made great progress over the years in financing both of those efficiently, expect to continue to be able to do so in the future. As one data point there, we recently entered into a $200 million sale leaseback agreement to finance our Charlotte IRC and our next three IRCs. After that, we intend to continue doing that going forward. I think we've been very pleased with the way that we've capitalized the business.
Making use of the most efficient capital markets to fund the assets that we have. We think in the long run that that is the best way to maximize return on equity for our shareholders. We'll continue doing that going forward. In terms of the way the business is positioned today, we have over $600 million in total liquidity resources, taking into consideration everything on our balance sheet, which we believe gives us significant flexibility.
Great. Appreciate it, Mark, Ernie.
Thanks. Have a good one.
Our next question will come from Nat Schindler with Bank of America Merrill Lynch. Please go ahead.
Hi, guys. Thanks for taking my question. Just wanted to dive into the retail unit growth projections that you're suggesting and compare it to kind of the market cohort data. Seems pretty conservative that you're going to slow down that dramatically, given that you're saying that your average 2013-2018 cohort was at 84%. Even if it, Nashville, and I forgot what was your other one in the second cohort. Even if both of those go the way of Atlanta, you're still shifting a huge number of markets will be in that kind of core, pretty high growth rate average that you're talking about, including, I would assume, the 2019 markets act in 2020 like the 2018 markets acted in 2019. If that's the case, how are you going to decelerate all the way down to well below your average for even your three-year oldest cohort?
Sure. Well, I think you're looking at the same data that we're looking at, so I think that's what's informing our guidance. I think the only layer that we would add on top of that is we did just have a very incredible growth year in both retail units and buying cars from customers. We think that we're set up very well to have another growth year again. Whenever you're growing that fast in total transactions, there's always going to be one or two things that pops up unexpectedly that we need to be ready for, and we're trying to make sure that we take all that into account. I think you're looking at the same data.
You would characterize the cohort the same way that we would, and then we're trying to roll that forward and make sure that we take into account just the operational realities of growing at this scale and speed.
Okay. Just to clarify, and I know looking at market numbers is somewhat useless because not all markets are the same, obviously. If you're going to go to 73%-75% coverage in this coming year, is it right to assume that you're going to add roughly the same number of markets, kind of in the 60 range that you added this year?
We think we'll add more markets now. We expect to end the year at approximately 250 markets this year.
Oh, I thought that was wrong.
Yeah. We think we'll add more markets. Those markets will not be super large markets. You can kind of calculate the average size of those markets given the amount of population that we expect to serve. I think markets are getting continually easier for us to add as our logistics fabric spreads out across the country. It gets easier and easier to add new markets. That's a number that I believe is in the 20s, then in the 40s and 60s. This year, we expect to be approximately 100 markets that we'll add. That's definitely getting easier for us.
Sorry about that last question. I just misinterpreted that slide on page 11 or misread it. Obviously, that data was there. Thank you.
All good. Thank you.
Our next question will come from Brad Erickson with Needham & Company. Go ahead.
Hi, thanks. Just a couple follow-ups. One, as you buy more of these cars from retail customers, obviously getting the nice tailwind to GPU. Given you kind of talk about passing along some of those savings or in the form of savings to your customers, is it fair to assume you're continuing to kind of incorporate those GPU tailwinds, at least partially into your prices, I guess making you even more competitive price-wise? That's the first one. Related to the shift to the two-shelf securitization you talked about, anything to be read into there in terms of just the mix of your financing business here in 2020 going through Ally versus other investors that's contemplated in your guide? Thanks.
Sure. I'll hit the first one, Mark why don't you hit number two. In terms of price competitiveness, if you look through 2019, we've got very similar discounts relative to other competitors in the marketplace to what we've had in the past. I think our pricing policies have been pretty similar across time, and so I think you can think of those benefits that we're getting from buying more cars from customers as largely flowing through and feeding GPU. At least that's what's happened in the last 12 months. Hopefully that's helpful color. Mark, why don't you jump in on number two.
Sure. Yeah. On the transition to the two-shelf securitization, we think that's a very natural evolution for our program. Obviously, we were really excited to get the securitization program up and running in 2019 and saw great success from that program in terms of the number of investors that we were able to reach and the impact on finance GPU. We think that the transition to two shelves is a natural evolution as we move to the long-term model. It expands your investor base. It gives investors greater liquidity. I think there's an opportunity for more efficient capital structures, all of those things enable lower cost of funds over time. In terms of mix with Ally, we still have a great relationship with Ally. We're selling loans to them, expect to continue selling loans to them.
I would almost think of that as unrelated to the move to two shelves, which we think is just the right next step in our evolution in finance monetization.
Our final question will come from Lee Krowl with B. Riley FBR. Please go ahead.
Great, guys. Thanks for sneaking me in here last. Two questions. First, just wanted to focus on other GPU and namely the auxiliary services you guys attach to sales. Curious how that's impacted as the inventory becomes more diverse, and especially as you're buying cars from customers and the age fluctuates relative to the last couple of years. My second question is, as we head into the seasonally stronger year, particularly with tax, curious on changes to strategy from a year-over-year perspective now that you have more capacity in place to handle the higher volumes from buying cars from customers. Thanks.
First question I'll take on ancillary product GPU. That's obviously a place that we've made nice progress. I think there can be some correlations between ancillary product GPU and vehicle mix. For example, if a car is within the manufacturer warranty, extended warranty or vehicle service contract might be a little less attractive, whereas if it's outside of manufacturer warranty, it might be a little bit more attractive. I think there can be some correlations between vehicle mix or other factors with ancillary product attachment rates. All of those, of course, would be incorporated into our guidance.
Then on the strategy point, heading into tax season and beyond in 2020, I don't think there's anything material to call out. I think we feel really good about how we're positioned right now, where we're coming in with a really big inventory that's very diverse, more diverse than any inventory we've ever had. I think what we're always trying to do is appropriately size the business to be able to handle the volume that we see today and hopefully be able to quickly flex to handle the volume that we see as we move through these high-volume times. I think we're really well-positioned. We're obviously growing extremely fast, so we'll be keeping a close eye on if we start to constrain in different places, and we'll be reacting just like we have in the past.
I think nothing material to call out there except that we're walking into tax season in good shape operationally.
Great. Thanks for taking my questions.
Thank you.
This will conclude our question and answer session. I would like to turn the conference back over to management for any closing remarks.
Thanks everyone for joining the call, and every bit as importantly, thank you so much to everyone on Team Carvana. This was another incredible year, and it's only possible because of how hard all of you work every single day. We really do appreciate it. We really do recognize it. Thank you so much. Have a good one.
The conference is now concluded. Thank you for attending today's presentation. You may now disconnect.