Description

Buying a used car has historically been a terrible experience; a fragmented, cottage industry delivering an opaque, time-intensive, and expensive transaction.  Carmax, a well-known used car retailer, has experienced remarkable success over the last two-plus decades by taking a more professional, standardized and customer-friendly approach to the market.  By offering a larger selection than competitors with an upfront price / no-haggle policy and transparent financing options, the company has experienced tremendous growth and delivered attractive returns to shareholders.  Additionally, due to the size (~40m used cars sold per year in the U.S.) and fragmented nature of the industry (top 100 players have 7% market share), Carmax still only has less than 2% market share and therefore a long runway for continued growth.

While I have admired and continue to admire Carmax, and think that they have many years of continued success ahead of them, this write-up will focus on a company that I think is even better positioned in the used car retail market: Carvana.  To be clear, I do not think that positive views on Carmax and Carvana are mutually exclusive, given the sheer size and fragmentation of the used car market.  However, while some people note Carmax’s dismissive / negative attitude towards Carvana as a reason to be skeptical of the latter, I happen to think that view likely has more to do with the fact that folks in Richmond still remember well how quickly Amazon leveled Circuit City (Carmax’s former parent company).  While I think this situation is different, and that both companies can co-exist and succeed independently, I think there is reason to assign some likelihood of institutional bias with regard to Carmax’s view of Carvana.  As a bit more of an aside, that bias / memory of Circuit City’s experience with Amazon does seem to be leading to Carmax being much more of a fast-follower of Carvana with respect to home delivery, online financing, online appraisals, etc., which I think is a testament to the viability of Carvana’s value proposition (I realize that sustainable economics are needed as well, and more on that later).

At just four years old, Carvana is smaller and lesser known, and certainly not as accepted by the investing public as Carmax (+80% of Carvana’s albeit small float is sold short).  Like Carmax, Carvana is also a used car retailer, however they operate entirely online.  Carvana operates in 37 markets (7 of which have vending machines) that cover approximately 35% of the U.S. population (for reference, Carmax operated 178 physical stores in 87 markets that cover roughly 55% of the U.S. population).  Through Carvana, a consumer can purchase a vehicle, with financing, in under ten minutes on the Carvana website or mobile app, and have that vehicle delivered to their door, often the very next day.  Additionally, Carvana has a 7-day no-questions-asked return policy, and indicative of their maniacal focus on the customer experience, their customer advocates will call you on the 6^th day to make sure that you want to keep the car; they are not trying to pull a fast one on you (the company does not disclose the percent of cars returned, but have indicated that it is a low single digit percent of volume).  I think Carvana is somewhat similar to Zappos; a disruptor in a category previously thought to be relatively impervious to e-commerce given the perceived try-before-you-buy nature of the product.  However, while that is the perceived nature of the product, 97% of vehicle purchases involve online research, 75% of people say they would consider purchasing a car online, and 52% of people test drive one or no vehicle prior to purchase.  Not to mention that Carvana is growing very quickly, a testament in itself.  Nearly every car dealer, including Carmax, while tell you that virtually 100% of their customers start their purchase process online.  Carmax CEO Bill Nash has stated that he would not be surprised if ~10% of the industry was purchasing cars online in the next several years, and I believe that’s why you see Carmax following Carvana into the areas I mentioned earlier.  Furthermore, the ease, transparency, and value (Carvana cars are priced ~$1,500 below Kelley Blue Book and Carmax) stands in stark contrast to the terrible customer experience that pervades the majority of the used car industry.  Carvana’s goal is “to offer customers what they want: lower prices, wider selection and simpler experiences.”  Unsurprisingly, customer reviews for the company are very strong (4.8 out of 5 star average, with over 10,000 reviews).

How can Carvana deliver this kind of value proposition and customer experience profitably?  They can do so by taking a clean-sheet-of-paper approach to the used car business in a period of time where e-commerce is ubiquitous.  The result is a radically different approach to the asset base and cost structure of the business relative to the rest of the industry. Founded two decades later than Carmax in a world where broadband connections are practically a basic need, businesses are being built around smartphones and mobile apps, and e-commerce is toppling brick & mortar retailers at a steady clip, Carvana saw an opportunity to serve the same large, fragmented and dissatisfied market as traditional used auto retailers, but with a radically different cost structure.  Carvana’s cost structure is much more centralized than a traditional dealer.  Rather than build individual dealerships in local markets at a capital cost of $10-20m each, where each dealership only holds a couple hundred cars for that market, employs a large staff and runs $6-8m of SG&A annually, Carvana can open a market at the cost of ~$500k for a small plot with a few employees, a few single car haulers, and a multi-car hauler to connect the market to their national inspection, reconditioning and logistics network.  The national / centralized network consists of four inspection and reconditioning centers (IRCs) across the country, and a logistics network where Carvana does nearly 100% of the transport between IRCs, between IRCs and markets, and the last-mile.  The company still uses 3^rd party transportation providers for a portion of the logistics between auctions and IRCs.  Regarding how differentiated / unique this network is, Carmax uses 3^rd party transportation for all moves of any considerable length (outside of a large format store to small format store move).  While the bulk of Carmax’s capital has gone into specific markets, the bulk of Carvana’s capital has gone into creating an impressive online experience and centralized inventory and logistics network.  As a result, Carvana’s PP&E per retail unit it not just falling, but is already lower than Carmax (which has been rising, as they’ve opened smaller format stores with less scale, etc.).

I believe that the difference in cost structure between Carvana and the rest of the industry is the main reason that investors underestimate the potential for Carvana’s unit economics.  Most investors assess Carvana’s potential by comparing their current unit economics to those of Carmax.  Given the radically different approach to their operations / asset base / cost structure that comparison is apples-to-oranges.  Carmax, while a very profitable business, has very little operating leverage.  Meanwhile, the whole point of the Carvana model is to have massive operating leverage.  I break Carvana and Carmax’s economics down into used retail, wholesale, and other / finance.  With regard to wholesale, I give Carvana little to no improvement in GPU from current levels given the fact that they do not have, nor have any intention to replicate Carmax’s impressive auction network.  Other / financing is more complicated; there are differences in both models, but more puts & takes rather than a clear benefit in favor of Carmax.  The key issues in comparing Carvana other / finance GPU to Carmax are: 1) Carmax has a nicely profitable capitive finance subsidiary and Carvana does not, 2) Carmax has to pay sub-prime lenders $1,000/unit to take a loan, while Carvana makes money on those loans given their significantly better LTVs as a result of their cheaper prices, and 3) the attachment rate and economics of other services like extended protection plans.  Currently Carvana makes about $700 of other / finance GPU.  Rather than compare to Carmax’s other / finance GPU to Carvana as a benchmark, I think the easiest / best way to assess Carvana’s future prospects are to break their current economics down into the finance (~$500 of the $700) and other (extended protection plans, etc. – the remaining $200 of the $700) and forecast each separately.  I take what I think is a fairly conservative route and say that over the next seven years (that is the time period over which I’m valuing Carvana / assessing a prospective IRR) the $700 of other / finance GPU will go to $900, and that all of that will come from extended protection plans, etc.  I think that method is fair and perhaps conservative given that Carmax currently generates ~$450 of GPU from extended protection plans, and that it sounds like there is room for Carvana to improve on the $500 of finance GPU given the attractiveness of their LTVs.  The last and most important piece of Carvana’s unit economics is the core used retail GPU.  For context, I think it’s important context that the company has already been experiencing significant scale economies; total GPU has gone from negative ~$200 in 2014 to over ~$1500 this year, and the core used retail piece of that has gone from negative ~$200 to ~$850 this year.  The biggest component of benefit to GPU going forward will likely be the effect of decreasing the company’s inventory days.  They are currently running at about 100 and believe that over the next several years they can get down to around 45.  This makes sense when you benchmark it to Carmax, especially given that Carvana’s model should by definition be more inventory efficient over time given they can sell a car across their entire footprint whereas Carmax is bound by local markets (mainly; they do transfer ~30% of their cars but I believe the majority of that is in between certain flagship / large format and satellite / small format stores that are not very far apart).  At a depreciation cost of $10 per day per vehicle, the inventory benefit to GPU over the next several years could be over $500.  Next is price; the company believes it has the ability to roll back some of its price discounting over the next several years and still be materially cheaper than competition and a great value proposition.  Note that if you add ~$200 in other / extended warranties, etc. GPU, ~$500 from inventory, and another ~$500 from price (they’d still be $1,000 cheaper than Kelly Blue Book and Carmax) to this year’s +$1,500, you are already nearly at the company’s mid-term target of $3,000.  Lastly, and importantly, that is before giving them any benefit from scale economies.  If you back out inventory effect on used retail GPU, you can see that the company has been generating several cents of GPU per incremental retail unit over the past several years; this makes sense given the centralized cost structure I’ve described.  This continues to be a significant opportunity going forward given the company is growing units in the triple digits, but you don’t even need much benefit from scale to get attractive unit economics.

Now that I’ve laid out why I believe you can underwrite a walk towards $2,500-$3,000 in total GPU over the next several years (I don’t even get all the way to the company’s $3,000 mid-term target to get an attractive return), I’ll try and cover the drivers and my expectations for the other key areas of the financial model.  Beginning with the top line, Carvana gave in their S-1 (which should be updated annually in the Ks) some useful information on market share curves by market, and detailed market case studies of Nashville and Atlanta in particular.  I’ve modeled their top line market by market using the population for each MSA and market share assumptions based on the aforementioned curves, the Atlanta and Nashville case studies, and assumptions of where the market share curves level out.  The curves are currently in the “straight up” part of the S curve, and assuming they level out in the ~5% range, or even lower, you can get some very attractive unit growth numbers.  I think it’s important to note that Carmax’s market shares in their mature markets is thought to be well over 5%.  Regarding market launches going forward, I assume the cadence slows in the years ahead and that they launch MSAs with populations at the median of the cities that I expect them to target (based on proximity to their inspection and reconditioning centers – i.e. I assume they don’t launch Boston, or New York for that matter, despite the fact that they have an IRC outside of Philadelphia).  On SG&A, regarding where their marketing / customer acquisition costs can level out, I use per unit assumptions that are based on current levels, the trajectories of Nashville and Atlanta as they’ve scaled, as well as Carmax figures and discussions with industry participants.  On the other line items I give no leverage to market occupancy costs, and I give some leverage (basically grow them at a slight discount to the overall growth rate of the business.  The output is an SG&A per unit figure of ~$1,000 over the next seven years, which I think seems very reasonable given that Carmax is ~$2,000 with a highly variable cost structure (huge fixed cost and staffing overhead in each market, as discussed).  Lastly on capex, the company can open a new market for $500k (again, vs $10-20m for a traditional dealership), which includes a small nondescript lot, a few employees, a few single car haulers for ~$75k each, and a multi-car hauler for ~$250k to connect the market to the centralized logistics and inventory network).  Regarding additional single car and multi-car haulers I use capacity break points of ~65 and ~230 cars per month, and for vending machines I use $5m of capital spend.  For valuation purposes, I use an exit multiple of 5x gross profit seven years out.  I use that multiple because it is where Carmax has traded, and given the leverage in Carvana vs Carmax (lower SG&A per unit at that point with better ability to scale both GPU and SG&A per unit), I believe Carvana should theoretically trade at a higher multiple of GPU.  I also think this multiple is conservative given that the implied EBITDA multiple is <10x, and the company at that point should cover ~55% of the U.S. population and still be growing double digits.  The IRR that I get is well into the double digits despite the fact that I model no share repurchases (a hallmark of Carmax and other growth companies with great unit economics) and thus the balance sheet is extremely inefficient / net cash at exit.

On why I think this opportunity exists:  In general I think Carvana is misunderstood because it is small and relatively illiquid.  More specifically I think Carvana is misunderstood because people attempt to compare the economics to those of Carmax, when the cost structures are entirely different.  In fact, the whole point of the company is that the cost structure is entirely different.  Carmax was founded in 1993, before the Internet was widespread, and buying anything online was out of the ordinary, let alone buying a big-ticket purchase like a car.  Given that the nature of retail had not yet been revolutionized by the Internet, Carmax approached the used car market with a similar cost structure as competitors.  Similar to how it sounds like Jeff Bezos has pushed Amazon employees not to focus on being / acting like a retailer, Carvana has taken an approach to their market with an explicit disregard for the traditional economics of used car retailing.

Regarding this recommendation and some comments I’ve seen on this site regarding value investments: I often see value in companies that can internally fund a growth runway that the market is currently implying has an extremely low probability of coming to fruition.  If that model is not encompassed in your definition of value, and therefore this idea is not of interest to you, I apologize.  Regardless, thank you for reading.

Risks:

1)  Execution: this is a very fast growing business that is launching nearly 20 new markets a year.  Even the best growth companies in the world (Amazon again comes to mind) have had significant execution stumbles.

2)  Hurricane noise: the recent hurricanes have the potential to introduce noise to near term results, however I look at this as temporary, and hurricanes have actually been well-documented as a net positive over several quarters for the used car retail industry.

3)  Capital: the company has significant cash on the balance sheet and an additional $50-100m of financeable assets and real estate on its balance sheet, but it is possible that the company will seek to raise additional capital at some point.  Given the unit economics and growth I am forecasting, the kind of limited capital raise that I think they might need, even at low stock prices, does not impact the prospective IRR significantly.

Note:  A lot has been said about the Garcia family as a risk.  I personally think this well overdone.  Ernie, along with the rest of the executive team, strikes me as highly intelligent /educated, highly motivated (he owns a significant amount of the business), and a generally nice / ethical guy.  The feedback I have had on him and the rest of the team from industry participants and former employees has been consistent with that very positive perception.

Catalyst

1)  Unit growth.

2)  Continued progress on GPU.

3)  Short covering.