[KMX – CarMax; CVNA – Carvana] Moat, Growth, and Competition

When I Google “used car salesman”, here are the first images that appear.

‘nuff said. It required an electronics and appliance retailer with no car selling experience to be the change that consumers wanted to see in the car buying experience.  In 1993, the now defunct Circuit City launched the first CarMax store, leveraging its inventory management expertise to replicate the hassle-free buying experience that consumers had gotten used to when shopping for so many other goods (indicative of its retail mindset, CarMax’s management refers to its lots as stores, not dealerships).

CarMax sets itself apart from traditional used car dealers in several ways.  First, whereas used car dealers mostly treat the sale of used cars as a loss leader to more profitable parts and service operations, CarMax focuses exclusively on used car sales. Second, it carries 4x to 5x more inventory than a typical car dealer, so customers can shop and compare models at a single location rather than laboring from lot to lot. Third, CarMax doles out a fixed dollar commission for every vehicle sold, incenting its staff to behave more as consultants pairing the right car to the right driver than as slick salesmen pushing the car with the highest price.  The retail price that the customer pays for a used car and the appraisal value that CarMax offers for the trade-in may change from day to day not based on the bargaining skills of a particular customer, but rather on local supply/demand conditions for the car model in question, conditions that are processed and analyzed by distinct groups operating outside the domain of the customer support guy whose sole job is to create a comfortable and informed buying experience for the end customer.

Also, CarMax appraises and purchases any car brought to the lot, regardless of whether that customer buys a CarMax vehicle.  By treating the buying and selling of vehicles as two separate processes and offering guaranteed cash offers for trade-ins, over time CarMax distinguished itself as a transparent, credible, and consistent venue for car sellers.  And today, a significant proportion (around half?) of the vehicles that CarMax sells to customers are sourced through such appraisals, which is significantly more profitable than sourcing cars through auctions.  The 50% of appraised and purchased cars that don’t meet retail standards – too old and too dilapidated to be refurbished at reasonable cost – are channeled to the wholesale market, where the company manages to clear 95% of vehicles compared to an industry average auction sales rate of just 60%-65%. And so, CarMax has a sourcing advantage relative to other dealers in the retail channel, plus the critical mass of owned vehicles to operate its own wholesale auctions [unlike the fragmented used retail market where the top 100 have just 7% share, auctions are dominated by two players, Manheim and KAR Auction Services, who lay claim to 70% of the North American market].  

These buying and selling streams in the auction and retail channels are important because of the data they bring. I’ve heard it said many times that CarMax’s primary point of competitive differentiation lies in its no-haggle one-price policy.  This isn’t quite right, as it doesn’t explain why CarMax has succeeded where others have failed – Auto Nation copied CarMax with fixed price, late model superstores in the late ’90s and took at $400mn+ charge shutting them down in 1999.  Auto Nation recently revived its fixed price effort but is stumbling on implementation.

The high fixed costs of retail and the low gross margins of used car sales are an unholy combination that leaves little margin for error with respect to how much you pay for inventory, how intelligently you price it, and how quickly you turn it.  There is no pricing power in the used car industry and no meaningful procurement leverage either, so really the only sustainable way to generate reasonable returns is to optimize the process of buying and selling cars using technology and data.  CarMax itself lost money in the first 7 years of its life trying to get this model right.

But this is where CarMax’s Circuit City lineage served it well.  From its earliest days, the company applied information technology and data-fed algorithms that retailers had long used to optimize the inventory and pricing of low margin, heterogeneous goods.  For instance, by placing an RFID tag on every car, CarMax can monitor how long each vehicle sits on the lot and how frequently it is taken out for test drives, which when combined with auction data provides granular insight into customer preferences local markets.  This information is used to alter the balance of makes, models, color, mileage, and price points. 

The following excerpt from a 2005 Harvard Business School case study describes how CarMax’s buying team recruits this data:

Each CarMax location had a team of experienced buyers who set the price offered to a customer for his vehicle.  Data provided to buyers allowed them to know how many retail vehicles had been purchased, the reconditioning costs on a given vehicle type, how many of those vehicles had actually sold, inventory turn rates on the vehicle type, and the profit margin on each vehicle. The buyer’s appraisal was also based on years of historical data, as well as more contemporary auction information.  Everything that the buyers did was tracked, monitored and compiled into a monthly report card.The other major source of CarMax inventory was external wholesale auctions, where CarMax buyers competed with other dealers for vehicles. Although some believed that CarMax overpaid for some vehicles, CarMax had detailed, historical information on how quickly they could expect to sell a given vehicle, and what their net margins on the vehicle would be. CarMax buyers typically went to auction with a list of vehicles, and how much to pay for each type of car…

After a car was bought, a corporate-based analytics team used a proprietary algorithm to calculate the probability of sale of each car in the system and helped determine the mark downs on cars that remained unsold at opening retail price points. Executive Vice President and CFO Keith Browning explained:

Suppose a Honda Odyssey minivan comes into the store, the buyer knows automatically to pay more for that because they’re so scarce in the resale market. Our analytics team helps the buyer understand when a car has a high probability of sale, and then their information helps the buyer set an initial price on that car based on the other cars that are in our system. We don’t always set the price correctly, but we have a markdown strategy if the price is set too high.

By closely aligning inventory with demand preferences the company clears 99% of the inventory it offers to retail at retail (and 95% of the wholesale vehicles it funnels to auctions)…moreover, by using data to inform its appraisal and auction buys in real time, it can sell its cars at a remarkably consistent price-cost gap.  Check out CarMax’s steady gross profits per used vehicle sold (GPU – gross profit per unit) compared to the volatile and declining GPUs of its peers:

If you were to stretch this chart all the way to 2002 for CarMax, you’d see more of the same.  During the last recession, when used car sales declined by a cumulative 16%, from 43mn vehicles in 2006 to 36mn in 2009, and average selling prices declined by ~20% (per management), CarMax’s GPU inched down by just 2% (vs. CarMax’s average used selling price decline of 6% during that period), significantly better than its peers.

There’s been some concern over the last few years that the impending flood of vehicles coming off lease will hit used car prices, driving lower gross profits per unit.  But this fear appears overstated for CarMax, which has managed its costs in line with revenue and maintained steady-to-growing GPUs over at least the last 15 years.  And to the extent that it widens the pricing gap between used and new cars, a surfeit of newly off-lease vehicles may even drive better used volumes.

Why don’t peers simply copy CarMax’s technology?  Well, first of all, like most enduring competitive advantages, this one – grounded in data and systems – has emerged through an iterative, decades-long process that has seen nearly 11mn vehicles sold and over 25mn vehicles appraised through 65mn store interactions.  It would be tough for a traditional used car dealer, including/especially any of the large publicly traded ones who have long been valued on near-term earnings, to slash their profits and invest huge sums of money building out a sophisticated inventory management system and, in the process, realign their organizational set-up to accommodate the systematized, high volume buying and selling processes required to make money on a predominantly fixed cost base.