Good afternoon, and welcome to the Carvana Fourth Quarter Earnings Conference Call. [Operator Instructions] Please note this event is being recorded. I would now like to turn the conference over to Mike Levin, Vice President of Investor Relations. Please go ahead

Thank you, Olson. Good afternoon, everyone, and thank you for joining us on Carvana's fourth quarter and full year 2017 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 and full year shareholder letter is also posted on the IR website. Joining me on the call today are Ernie Garcia, Chief Executive Officer; and Mark Jenkins, Chief Financial Officer. Before we start, I would like to remind you that the following discussion contains forward-looking statements within the meaning of the federal securities laws including, but not limited to, Carvana's market opportunities and future financial results that involve risks and uncertainties that may cause actual results to differ materially from those discussed here. A detailed discussion of the material factors that cause actual results to differ from forward-looking statements can be found in the Risk Factors section of Carvana's annual report on Form 10-K for 2017. 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. And now, with that said, I'd like to turn the call over to Ernie Garcia. Ernie?

Thanks, Mike, and thanks, everyone, for joining the call. As we have in prior quarters, we put a shareholder letter out on the Investor Relations site. We put a lot of detailed information in the letter relating to both the fourth quarter and the year as a whole, so I'll spend most of my time today focusing on key milestones and trends in our business. Let me start in the obvious place. There's admittedly some noise in our results. We missed our own guidance in unit sales in the fourth quarter. We also grew revenue 148% year-over-year, our highest growth rate ever as a public company, which begs the question, how did that combination of facts happen? The answer is that we had a great fourth quarter by any reasonable measure. We also had unexpectedly weak results from our Cyber Monday promotion that surprised us, and we failed to sufficiently factor the unexpected into our guidance. That noise was completely self-inflicted. That said, it is noise. So what are we excited about as we head into 2018? In the shareholder letter, we included several graphs that we believe highlight the most important financial measures of the business. They include units, revenue, GPU, number of markets, number of vending machines and EBITDA margin. Examining these numbers annually is helpful because it provides a bigger perspective that makes the Carvana story very clear. Please take a moment to look over these graphs. In addition, we include our cohort market penetration curves. We also really like the story these curves tell. Our growth continues to be significant and consistent across markets. Our most mature cohort, the 2013 cohort, reached a market penetration of 1.54% on growth of 44% in the fourth quarter year-over-year. We don't know how high our market penetration can ultimately be but even if you assume that 44% growth rate drops to 0 from here, the penetration we have already achieved in that cohort would enable us to get to a scale that is over 13x where we ended 2017 if we can get nationwide penetration to the same level. We are just getting started. Carvana is now a 5-year-old company. In that short time, we have established ourselves as the leaders in automotive e-commerce. Our customers can buy a car in minutes on our website while getting approved for financing, pricing their trade-in and signing contracts without ever leaving their couch. They can have their car delivered as soon as the next day through our in-house logistics network, with the car arriving at their door accompanied by a Carvana customer advocate who shares their enthusiasm for the moment. This offering is powerful and difficult to replicate as it requires expertise in many areas. The numbers reflect this power and uniqueness. All of this has been accompanied by significant GPU growth and significant operating leverage. Over the last 3 years, we've grown GPU from negative $200 per unit to over $1,500 per unit, and we have improved our EBITDA margin from negative 32% to negative 17%. We've achieved this while rapidly scaling, evidenced by a cumulative annualized growth rate of over 175% and 140% in units and markets respectively over that period. We view this as a significant accomplishment with few precedents in either the public or private markets. Accomplishing all of this so quickly is a testament to 4 things: the size of our opportunity; the quality of product we have put in front of our customers; the scalability of our model; and most importantly, the caliber of people who have made the Carvana mission their mission. These graphs extrapolate to achieving massive scale in straightforward ways. So what stands in our way? We believe the biggest hurdle we face is execution risk. We need to continue scaling sourcing cars for our customers, certifying them, transporting them through our logistics network and delivering them to our customers' doors and, most importantly, wowing our customers every step along the way. Each step carries risk, and we think a useful way to examine these risks is to separate them into 2 buckets: the operational bucket and the customer experience bucket. The operational bucket is a big one. The good news is that this risk is -- mostly impacts the time it takes us to achieve any given milestone more so than the ability to ultimately hit that milestone. We're very focused on anticipating and mitigating operational risks and have an incredible team with strong leadership in each operational area, which gives us confidence in our ability to manage this risk going forward. This confidence is underscored by the fact that we opened 23 markets in 2017 and have already opened 10 more so far in 2018. We also opened our first Carvana build inspection center and opened 5 vending machines, the most in our history. That all came together with the delivery of over 44,000 cars to happy customers this year. Second is the customer experience bucket. This reduces to a simple question: can we continue to deliver and improve upon the customer experience we pioneered at the next order of magnitude? In our opinion, this is the most important risk we face because it not only impacts the time it takes to achieve any given milestone but also dictates which milestones are ultimately achievable. This is a measure, we have fared very well on to date. We now have reviews from over 14,000 customers with an average rating of 4.7 stars. 10,000 of those reviews came in the last year alone. Given these risks and the sheer size of the opportunity we face, our strategic plan is to be aggressive while always keeping the customer experience at the center of every decision we make. Opening markets quickly kicks off key network effects in inventory, marketing and brand, but also generates additional execution and customer experience risks that we must be mindful of. This reality puts our goals of aggressively seizing this opportunity and delivering the best customer experience available to our customers in conflict to some degree. That conflict is resolved by the quality of our operations. As long as we continue to execute at the level that enabled us to grow revenue by 135%, grow markets from 21 to 44 and to deliver customer experiences that earned us those 14,000 4.7-star reviews, we will continue to accelerate in this exciting opportunity so our customers, team members and shareholders benefit from the positive feedback of growth inherent in our model. We know that the triple-digit growth tapped into this tremendous opportunity will not always be a straight line, but we believe we will achieve our goals as long as we keep our eyes on our mission to change the way people buy cars. We're entering 2018 proud of our accomplishments, but not satisfied. We also believe the opportunity we have is incredibly unique, and we feel privileged to be the team tackling it. We are up to the challenge. Mark?

Thank you, Ernie, and thank you all for joining us today. Unless otherwise noted, all comparisons are on a year-over-year basis. We are pleased to report another quarter of triple-digit growth in both retail unit sales and revenue. Retail units sold totaled 13,517 in Q4, an increase of 141% and an acceleration from Q3. Total revenue was $265.1 million, an increase of 148%. We expect retail units sold of 17,000 to 18,500 in Q1 an increase of 104% to 122%. We expect the timing of IRS refunds this year to be similar to last year. Total gross profit in Q4 was $21.9 million, and total gross profit per unit in Q4 was $1,619. For the year, total gross profit per unit was $1,539, an increase of $516. Our growth in GPU in 2017 was broad-based, including gains in purchasing and pricing, inventory management, reconditioning, financing and new products. We expect total GPU to be $1,650 to $1,750 in the first quarter followed by significant increase in Q2. Our outlook for Q1 GPU is lower than we normally would have expected in Q1 due to several factors, including higher-than-normal wholesale prices on purchases in Q4 after Hurricane Harvey, followed by accelerated depreciation rates in December as well as due to a sharp rise in short-term interest rates from Q4 through present, which impacts financing gross profit. We view these effects as largely transitory and expect a record quarter for total GPU in Q2 as we march toward our midterm goal of $3,000 per unit. In addition to increasing GPU in 2017, we also successfully reduced our average days of sale through our inventory strategy. We exited the year at 72 days in Q4, our lowest quarter ever, and expect further gains in 2018, particularly in the second and third quarters. We expect reduced average days to sale to contribute to higher total GPU in 2018. Our outlook for the full year is $1,950 to $2,150, an increase at the midpoint of $511. EBITDA margin was negative 15.5% in Q4, a slight improvement from the prior quarter. As with GPU, we expect EBITDA margin to improve in Q1 and then to improve significantly in Q2. In terms of our full year outlook, we expect significant growth in retail units sold, totaling 89,000 to 93,000 for the year, up from 44,252 in 2017. We expect to open 30 to 40 markets in 2018, reflecting an acceleration of market openings relative to the 23 we opened in 2017. As we move into 2018, we expect to continue to grow retail unit sales, open new markets and drive operating efficiencies to improve GPU and EBITDA margin. We expect OpEx to increase in dollar terms as we expand, but to decrease as a percent of revenue. We expect very modest increases in fully exchanged share count in Q1. You should use approximately 137 million shares on a fully exchanged basis. We ended the year with $172.7 million in cash and equivalents. In the fourth quarter, we entered into a Master Sale-Leaseback Agreement that provides for the sale-leaseback of up to $75 million of existing and future vending machines and IRC construction projects. We sold and leased back $19.2 million of vending machines under this agreement in Q4 and intend to sell and lease back additional assets over time. As we discussed previously, we believe this demonstrates our ability to use our balance sheet to provide additional financial flexibility as we continue our rapid growth. We also increased our financial and operating flexibility in Q4 by issuing $100 million of convertible preferred stock to Dundon Capital Partners. This preferred stock pays an annual dividend of 5.5% and is convertible to common shares at the conversion price of $19.6945. Due to the rapid increase in our stock price following the announcement but prior to the closing of the deal, this issuance gave rise to a beneficial conversion feature under GAAP, which increased the loss attributable to class A common stockholders by $1.237 million in Q4 and will increase the loss attributable to Class A common stockholders by $1.38 million in Q1 2018. In our first year as a public company, we made significant progress toward achieving our financial goals. As we look toward 2018, we intend to continue to expand our footprint, rapidly grow sales, increase GPU and demonstrate operating leverage. Thank you all for your attention. We'll now take questions.

[Operator Instructions] And our first question will come from David Lim with Wells Fargo.

Ernie and Mark, I don't know if you guys have this metric, but I'm sort of wondering, if you look at Q4 2016, I think you guys had 21 markets then. How did those markets sort of fare in Q4 of 2017 sort of like on a same-store sales? I guess this would be like a same-store geographic metric.

So I think you can kind of see that in the cohorts. I mean what I think I would do is I would look at the cohort [ personally ] put in the shareholder letter and basically look at all the curves that came prior to 2017. I think you can get a pretty good sense. So I think that's where we would point you. We're not currently giving a same-store sales metric, but those curves have been remarkably consistent, hopefully you can see that in those curves.

Got you. And then my follow-up question would be I think you said a $500 improvement in GPU, half of it coming from inventory. Can you sort of parse out the other $250, where the other $250 benefit is coming from?

And David, are you asking about 2017 over 2016?

I'm sorry, the 20 -- I think it was the 2018 guidance on GPUs. I think it was a $500 improvement over '17. Half of it's coming from inventory.

Right. Yes, right, yes. I'll take that one. So in terms of our guidance for 2018, I think the other half of the $500 total increase is going to be broad-based. I think we expect to see gains in wholesale gross profit per unit as well as the various other buckets, financing as well as our ancillary products. So I would just think of that $250 as being broken up across the rest of the transaction, but we expect improvements in all areas.

Yes. And David, I think that's' been pretty consistent across '17, in our expectation into '18 and even forward from there. I think as we said in the past, on our path to $3,000, we expect about half of that to come from products, about half of that to come from scale. We've been roughly eating into those buckets at kind of the same speed as we've moved through the growth we've seen over the last 1.5 years. And we expect to continue that at roughly a similar speed going forward.

Your next question is from Sharon Zackfia with William Blair

I think in the shareholder letter, you talked about a test that you did in marketing in Charlotte and Nashville that didn't work very well. So I'm just curious what that test was, what you learned from that about how your marketing effectiveness could evolve going forward. And then secondarily, on the vending machines. Did you talk about how many vending machines you're planning on opening in 2018? I didn't see that.

Yes. So I'll let Mark take that. So first of all, we're running marketing tests all the time. I think the nice thing about having 44 markets opened now is we're able to test all kinds of different treatments, whether it's different creative mixes, different channel mixes, different ways we present to the market. So we're testing different things all the time. I think we did run a fairly substantial test in both Charlotte and Nashville, which are the 2 markets that are under 2014 cohort, where we actually changed the creative on TV and we also made some changes to a number of channels. We ran that test for around 9 months starting at the end of Q1 roughly through the end of this year, and then kind of reverted back to our standard marketing mix at the end of 2017. That did not work out great. Normally, I think we wouldn't call out specific test results for marketing. I -- we're doing that because you can see the 2014 cohort, it kind of looks like it folds over and stops growing. And that's basically the results of those marketing tests. And then it's the blend of Charlotte, which is one of a small handful of markets that's been below Atlanta from the beginning, and then Nashville kind of being blended together. So we just want to point that out that that's driven by our marketing tests.

And then on the numbers ...

And then on the vending machines, Sharon, we are building vending machines. We have several vending machines in progress, some of which are visible to varying degrees. We're not providing specific guidance on vending machines for 2018.

The next question comes from James Albertine with Consumer Edge.

Wanted to ask, I think last quarter you provided a metric on customer demographics. You talked about millennials, I believe, or a certain percentage. Maybe I assumed it was millennials, but you talked about a percentage of customers that were buying from beginning to end on their mobile phones. I was wondering if you could update that number for us for the fourth quarter?

So I think as it relates to both millennials and kind of mobile trends, the trends are very consistent, they both continue to trend upward fairly slowly but consistently. I don't think we have any additional numbers to give you on this call. The numbers from last quarter's call are still pretty consistent.

Okay, great. And as a quick follow-up, if I may, as we think about advertising costs that you're leveraging a lot of your nationwide branding initiatives, just wondering if you sort of have a good sense now with some marketing tests perhaps that have failed and now you're sort of retracing and going in other directions? Do you have a pretty good sense now of the base? And it's only a question of sort of opening new markets and leveraging with unit volume over time? Or should we expect an additional sort of step function change in advertising spend looking ahead to 2018?

Yes. So I -- the first thing I would do is I would point to those cohort CAC curves. I think those are also very, very consistent, just like the market share curves. You can see them all kind of trending down in a very tight spread to one another. Yes, I believe in the fourth quarter, Atlanta got down to $440, and the rest of the curves are on a very similar path. So I think in terms of getting the entire company down to $440, that basically is a matter of getting all those other cohorts to catch up to Atlanta. Atlanta, despite growing 44% in the fourth quarter, also lowered its marketing expense, its CAC by about 20% year-over-year. I think those continued reductions are the results of continued learnings that come from all these different marketing tests that we run as well as word of mouth in the market. And interestingly, I think the 5-year transaction cycle in automotive, we have -- we're a 5-year-old company. And so we've been in Atlanta now for one automotive purchase cycle basically. And we've been advertising on TV for about 4 of the 5 years that we've been around. So I think there's still reason to believe that there will be strong valuable momentum in those curves over time. I think the unit economics work great as long as we get all the markets down to where Atlanta is today. But I think there's reason to believe that we can push even further down. There's a lot of momentum in those curves. And then -- go ahead.

And Jamie, to answer the second part of your question. We do think of advertising spend as being highly correlated with the number of markets in which we operate. So that pattern has been consistent over time. There is some variability in that. There's some seasonality, there's also some variability in cross markets in advertising intensity. But as a good approximation, our advertising spend is highly correlated with our number of markets.

Our next question comes from Nat Schindler with Bank of America Merrill Lynch

Just a couple of questions, but so just looking at the cohort analysis -- and thank you for providing them, that's really great detail. But if you look at the 2014 cohort, which I believe is just 2 cities, that's Nashville and, I believe, Charlotte, they took a tick up. Is that because of the -- in both -- the tick down in market share as well as a tick up in marketing cost per vehicle that's uncharacteristic of all the other markets. Is that because of the marketing plan that just didn't work the way you expected it to in those markets? And do you think -- or is there something more seriously going on in those markets? And then I have a couple of others.

Yes. So I think we definitely would attribute that to the marketing tests and then also just as a blend of those 2 markets, as you said, it's Charlotte and Nashville. Charlotte has always grown steadily but been lower than Atlanta. And that's kind of pulling the whole curve down. And then the marketing test I think is what kind of drives that shape. There's some other methodological reasons. We only show the cohort curves, you can only see points where every market has made it through that quarter in its life. So for example, the last data point is Nashville's third quarter and Charlotte's fourth quarter. And so you don't see Nashville's fourth quarter in there. Both of those markets are now growing market penetration again, so you just kind of can't see that in the underlying curve. And what I would say, I do think that's driven by the marketing tests. And I think the more important thing there is just those curves are very, very consistent. I think what's more important than the consistency of those curves is the driver of that consistency which is our business model is very scalable. It's centralized and so it gets -- it's very easy for us to roll out markets quickly and deliver a customer experience that consistently leads those curves.

Okay. That folds back into my next question, and I just have 2 more. Sorry, everyone. But so you're going to roll out as many -- you're going to double your markets again this year, and that's adding incrementally twice as many as you added last year. That's a big move. Is it getting easier to open markets at this point?

Yes. But I think there's been a number of operational improvements that we've made. So in the last call, we talked about kind of our hub in a box, which is an innovation that we kind of put together to enable us to open up markets more quickly. I think as our logistics networks spread across the country, it gets easier to fill in markets. The hardest part of expansion for us today is really expanding that logistics network more so than even adding additional markets along that logistics network. So now that, that logistics network is broadening, it's getting much easier to open markets.

Great. And one last question. I'm looking on your site now, and you have over 11,000 cars available. That's a lot more than your historical range of around 7,500. I know there is a seasonal aspect to this. Can you go over some of the seasonal aspect and also talk about whether or not there is going to be new heightened levels of inventory available as you get up to market the 80-or-so markets?

Yes. So I would say, in general, we held our inventory fairly flat last year. As we kind of talked about throughout the year, a major goal of last year was to drive down our turn time. We did that pretty successfully in the fourth quarter, our turn time was 72 days off of 97 days in the third quarter. So that's a tremendous amount of progress in a single quarter and I think it's testament to the real leverage inherent in the model and just kind of being able to sit all shelves at once I think speaks to our ability to continue that curve. We also tend to grow inventory heading out of the fourth quarter and into the first 2 quarters of the year where you get tax rebate season and tend to see kind of larger discontinuous growth. And so we grew our inventory a bit. I think you should expect the shape this year to be roughly similar to last year, relatively flat, and then growing again toward the end of the year, heading into next year's tax season.

Okay. And then just as a final last question, as you -- obviously this goes beyond your 2018 guidance. So -- but as you look at your cash position now after the preferred offering in December and how you're getting the marginal -- the improving declines in EBITDA as a margin basis, obviously trending towards breakeven, do you make it to breakeven, do you believe, on existing cash? Or do you believe there could be another raise?

I can take that one. So just to walk through the basic stats there, we had $173 million on the balance sheet at the end of 2017. We also had $55 million outstanding with our recent Master Sale-Leaseback agreement to be able to sell over vending machines and other real estate assets. When you combine those 2 things, plus our demonstrated ability to finance our balance sheet, including our floor plan line and all of our success financing haulers from real estate, we're feeling very good about where we are from a cash position.

Yes, and as it relates to raising money, I think we'll handle that the same way we always have. We're going to continue to monitor the business. We'll monitor opportunities for investment in the business. And we'll keep an eye on the capital markets and how we can go out and finance our balance sheet. And we'll make the right decision for the business and our shareholders when the time is right.

Our next question is from Sameet Sinha with B. Riley FBR.

Can you elaborate on what exactly happened during the Cyber Monday and the hurricane promotions, what exactly went wrong and talk about the learnings from there? Second question is, if you look at the GPU guidance, I mean you're basically implying that GPU at the midpoint, your guidance implies about a $520 improvement, half of that from used cars, probably $260. This is a slowdown from the $400 or so that you did this year. So from that $250 improvement, should we expect that to go down below $250 in '19? I'm just trying to get a sense of where this could peak.

I'll take the first part. And then Mark, why don't you jump in on the second? So I think the Cyber Monday promotion, we basically merchandised the Cyber Monday promotion and marketed a little differently than we did in years past. And I think that clearly was not as effective. We probably gave up something on the order of 1,000 units if we just kind of extrapolate what we would have expected. And the Cyber Monday promotion was similar year-over-year growth to what we got across the rest of the business versus what we actually got. So I think we made a couple of merchandising marketing choices that were just less effective there, I think that's what drove it. Now I think the good news is if you look at Atlanta, we called out that it grew 44% year-over-year in terms of market share. If you take out the promotion, which was soft across all the markets but extra soft in Atlanta, Atlanta actually grew at 58% year-over-year, up from 53% last year, excluding promotion sales. So Atlanta actually accelerated. So I do think the business performed very well in the fourth quarter. And then we did not handle the Cyber Monday promotion as well as we have in the past, and that ended up causing us to come a little short.

Great. And then I'll hit the GPU point. So I think that one thing to keep in mind as it relates to our 2018 guidance on total GPU as well as our breakdown is one thing that's impacting that is Q1 GPU in total and also retail GPU are lower than we normally would have expected, given some of the factors that I described, particularly the dynamics -- the wholesale pricing dynamics that we saw in Q4 and then the corresponding depreciation rates in December leading into Q1. And so I think that's one factor to keep in mind. In terms of the relationship during 2018 increases and 2019 increases, I wouldn't infer anything particular about our 2019 from the guidance that we've given for 2018. I think the way we think about this in a big picture sense is we're marching toward our $3,000 midterm goal. All of the key drivers or walk to the midterm goal are intact, and we feel like we're just marching down that path.

Our next question is from Steve Dyer with Craig-Hallum

You talked about the cohorts. And as you pointed out, Ernie, they're trending very -- pretty much the same on every one very consistently. On one hand, I would expect maybe they would -- the newer ones would trend a little bit more strongly, just given the ability to leverage advertising, et cetera. On the other hand, you obviously have some of your more appealing markets behind you. How do you sort of think about the cohorts in the markets you roll out from here? Is this sort of the benchmark you hope to set? Or do you think you can sort of improve on that?

I think as long as we continue to roll them out similarly to how they've rolled out in the past, I think we're going to be very happy. I think some of the drivers of how any given market performs, as we've spoken about in the past, in general, smaller markets ramp even faster than bigger markets. In general, markets that we are launching today are ramping faster than markets that we launched in years past, in general, Mark, they're closer to preexisting markets, have tended to ramp faster. And we've attributed that to more preexisting brand awareness. So I think there's a couple of things there that could lead us to be hopeful for even faster growth in the future. But I think as long as it's consistent, we're going to be very, very happy. And that's kind of at the market level. At the cohort level, there tends to be a blending of many of those factors, which is also I think what drives so much of the consistency. In most of those cohorts, there's -- well, I guess the more recent cohorts, there's 20 markets, and then there's 10 markets in the one before that, et cetera. So I think that's part of what's driving all the consistency as well as that all the other variation kind of gets massaged out in terms of size of market and proximity to nearby markets.

And then Mark, I haven't had the chance to read the K yet. But can you remind us what you're paying on your floor plan line and sort of how we should think about that this year in light of interest rate movements?

Sure. So the line does have a variable interest rate. Right now, that interest rate is just over 5% on the floor plan line. And then I think the way -- we expect these floor plan line going forward similar to how we used it in the past. We used that to pay inventory, used car inventory. And to the extent that, that interest rates move, our interest expense on that line will move. and I think we actually have a touch of commentary in the K that helps you quantify that. But that should be reasonably understandable from what you see in the K.

Our next question is from Seth Basham with Wedbush Securities.

My first question is around inventory management. As you guys entered the fourth quarter, I would assume that you expected similar year-over-year tax refund seasonality. However, it seems like your inventory is a little bit imbalanced, a little bit heavy here with more depreciation than you are anticipating. What are you guys going to do to better manage your inventory through these seasonal effects going forward?

I mean, I think managing inventory, I think, is a function of projecting demand and then buying the right cars. And we're going to do the best job that we can to project demand by extrapolating those cohort curves out. We got a great sense of all the different seasonal factors that drive automotive retail. I think that understanding is growing with every passing year as we get more information. And then I think we have a really good sense of exactly which cars our customers want because we have all the information of them clicking around on the website that helps to inform which cars we want to buy and put in front of them. So I think that's sort of a multipart question, where there's a question about which cars we're buying. I think we feel very good about that. And I think that as we get more data, we continue to get better. There's a question about how many cars are we buying. That's a question of forecasting demand. And to the extent we misforecast here and there, there -- we'll be a little heavy or a little light in inventory sometimes, but we're going to do the best job that we can.

Got it. Just so I'm clear, so the forecasting demand was off. You're not concerned or anything about the inventory and the quality of inventory right now?

No. I mean, one way to talk about that particular concept, Seth, is in the average days of sale. So we did say is we expect average days of sales to be declining in Q2 and Q3. We expect to be roughly in line with Q4 and Q1. And so I think we're happy with continued progress going into 2018 on average days of sale, which I think is a pretty good metric for how we're managing inventory size. I think the thing that we discussed as it relates to Q1 '18 on inventory is there was a unique wholesale market dynamic in Q4, given the hurricanes. I think that made depreciation rates and the timing of accelerated depreciation rates look different than other years. And I think that causes some -- I don't know if I'd say difficulty, but it just makes forecasting depreciation rates different when you have a major market shock like that. And so that's the way kind of I would think about -- that's the way I would think about the inventory management question. But I would just reiterate again, we're very happy with the progress we made on average days of sale through Q4. And I think we're optimistic about the continued progress we'll make throughout this year.

I really appreciate all the cohort data that you guys provided. Just one question on the CAC cohort data. CAC excludes number of advertising costs. Approximately how much of your advertising costs in total are excluded from CAC?

So the -- I think what you are referring to is when we started national cable, a lot of our national cable is hitting markets that are not in any of these cohorts because these cohorts only include markets that we've launched with the full Carvana delivery experience. And so the primary advertising expense is not included in these cohort curves. It's again, some national advertising that's hitting other markets that we haven't launched yet.

Understood. Could you provide an approximate amount or not?

Yes, it will be in the K.

Our next question comes from Ron Josey with JMP Securities.

So I guess a few questions, and maybe more higher level for Ernie and Mark and around 1Q guidance. It suggests a pretty big decel in growth overall. And understanding, we're now 2/3 into the quarter, is this all because of primarily the IRS trends you highlighted sort of similar to 2017, or something else? And then as it relates to 2018 guidance for units, you're potentially doubling your markets again this year. Understood it's easier to ramp new markets. So why shouldn't we see growth in units sold at least be steady, maybe even potentially accelerate? And then the last question, the third one, sorry guys, just in terms of the marketing spend. With 4Q units coming in below expectations, and Mark, your comments on cable TV advertising, understood the impact from Cyber Monday, et cetera, but do you think that as you grow more national advertising, which I think is certainly more efficient, that impacts your local exposure? And so therefore, you might need to just revert back to local over national? Hopefully that made sense.

I'll start swinging at the first one, and why don't you come back in on marketing? So I think Q1 first kind of like outlined how Q1 unfolds. In general, January and February are much like November and December. And then tax refunds tend to come in at the very end of February, early March. And that tends to be where a lot of the discontinuous growth comes from. We tend, in a single month, face thousands and thousands of units of growth at our current kind of scale. And so I think what we're guiding to in Q1 it's recognizing that there's a lot of scale coming very, very quickly. And our ability to handle that in a single month, it's going to be difficult for us to do that. And so I think it has more to do with kind of our expectation on the supply side than on the demand side as it relates to Q1. So that's what I would kind of say for a decel heading into Q1. It just -- we see thousands of units of growth coming very, very quickly there, and that creates some operational difficulties. Now your next question I think was about all the incremental markets and how does that impact what we expect for this year in 2018. So a couple of things. I mean, first of all, what you're saying is exactly right. The way that we provide guidance, I mean, there's basically 3 major inputs to that. It's what do the cohort curves look like, how many markets are we opening and how many vending machines are we going to launch on top of that because we have an expectation of how that works, and that basically drives our unit expectations. If you look at those cohort curves, you can see that they're exponential and they're relatively flat early on. Each of the last 2 years, we, roughly speaking, doubled our markets which, proportionately, is similar to what we expected this year. And in the last 2 years, our sales in those years that came from those newly launched markets were between 8% and 10% of total sales for the full year. So those markets, they definitely contribute, but they start to contribute really in years 2 and 3 more so than they contribute in year 1. And that's all taking into account, I think if you look at the cohort curves and extrapolate that yourself, you'll probably arrive at somewhat similar numbers.

And then Ron, on your point about national advertising and the relationship, being able to test marketing creative or other marketing mix, I think that's a good point. I do think as we scale national advertising, we'll have to be thoughtful about how we're testing new creative, new messaging. We will look to do that testing at the local level, generally speaking, before rolling out to a national audience. But that's something we'll certainly be thinking about and taking into consideration of our learnings from national and the rest of our experience here.

Our next question is from Gary Prestopino with Barrington Research.

Was there any changes in the amount of cars that were financed as well as warrantied and with GAP insurance in the quarter that you could call out year-over-year?

Gary, those attachment rates have largely been stable. So no meaningful change to the call out. One thing we've called out on previous calls is that we were very pleased following the introduction of GAP waiver coverage at the end of Q2. We're very pleased with the take-up we saw there from our customers. And that was a place where we saw steady increases in attachment as we rolled that out to more and more customers following testing and the like. And so I think GAP was a positive -- very positive experience for us. Other major attachment rates are roughly flat.

And where are you in terms of the Ally funding commitment? It's up to $2 billion, but how much have they taken on that?

Sure. We had significant capacity on the Ally commitment over $1 billion.

Okay, over $1 billion. And then just lastly, Ernie, just kind of an industry question here, we're kind of in an environment where rates are going up. Based on your experience in the business, where does that pain point start to become an issue for consumers, as rates go up, in the purchase of a vehicle?

I think that's hard to say. I think we're starting to hear a little bit of noise -- of softness in the market that may be partially driven by rates. I don't think we've really seen that in our direct results yet. I think so far, our growth has been fairly stable and predictable. I think -- I do think it's probably more likely than not that we're closer to the top of the cycle than the bottom. And so yes, I think there very well could be some degree of macroeconomic headwinds that face the market over the next couple of years. But we're not seeing anything specific that we would say merits a call out today.

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

Thanks again to everyone for taking the time to listen to our call. And thanks to everyone on team Carvana who's adopted our mission to change the way people buy cars. We're tackling a mountain here. The progress we are making is remarkable, and it's only happening because of all the passion and energy you bring with you every day. Thank you. In closing, I'd like to reiterate the points that I believe are most important to Carvana's story. We are expecting our fifth straight year of triple-digit growth as a company. Our cohort market penetration curves cut a clear path to significant growth beyond that. Our most mature cohort, the 2013 cohort, grew at 44% year-over-year in the fourth quarter and actually accelerated its growth to -- or versus 2016 when controlling for Cyber Monday sales. We plan to open 30 to 40 markets this year. We continue to make significant progress in our GPU march with over $500 of progress in 2017 and another $500 expected in 2018. Our cohort CAC curves shows clear path to significant leverage, and there's clearly lots of additional leverage elsewhere in the business. We expect to cut our EBITDA losses by 8% to 10% of revenue this year despite the investment in new markets. Our brand is defining what automotive e-commerce means to consumers, and replicating that experience is proving to be incredibly difficult for those that are trying. And finally, all of this is happening against the backdrop of a $1 trillion market. Thank you again for your time. We'll see you on the road.

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