KAR Auction Services says system on its Adesa.com marketplace tailors suggestions so dealers buy used cars that sell faster, bring more profit
By Jared Council, Wall Street Journal PRO writer
December 18, 2019
Vehicle auctioneer KAR Auction Services Inc. is using an artificial-intelligence-based recommendation engine to show dealers visiting its online marketplace which used cars are likely to sell fastest and generate the highest profit margins.
The Carmel, Ind., company added the feature to its flagship Adesa.com marketplace this year. The recommendation engine analyzes vehicle information, such as condition and mileage; market data, such as which makes and models are popular in a given location; and dealer history, such as which cars have tended to sell quickly at a specific dealership.
Based on those factors, the machine-learning technology comes up with recommendations and explains the decision. For instance, if a black Chevy Blazer is recommended, the system will display as many as 13 contributing factors, such as “Recent success: A similar vehicle has recently moved off your lot, it looks like you sell these well,” or “Pricing: The cost of the vehicle is 84% of the average market list price.”
KAR, which had revenue of $3.8 billion last year, runs online auctions through Adesa.com and smaller marketplaces. It also has 74 physical-auction locations in the U.S., Canada and Mexico.
The recommendation engine, built in house, is free for dealers to use. KAR makes money by charging auction fees to both buyers and sellers, as well as through other services. Auction fees, which depend largely on the value of the vehicle, can range from about $100 to more than $1,000.
The U.S. used-car market is growing faster than the market for new vehicles. Roughly 17.1 million new cars are projected to be sold in 2019, according to car-shopping site Edmunds.com, down 1.2% from 2018. Meanwhile, Edmunds forecasts that used-car sales will reach a record 41 million units this year, up 2% from last year.
Many dealers want to capitalize on the trend, but a key challenge is sifting through thousands of listings to find vehicles likely to sell quickly and profitably. Dealers have various sources for buying used cars, including from consumers via trade-ins. But many come from businesses selling them on marketplaces such as Adesa.com, which has roughly 70,000 listings at any given time.
“Dealers invest a lot of time in trying to find the right vehicles,” said Peter Kelly, KAR’s president. “So these tools that assist our buyers…take out a lot of the guesswork and they save time.”
The KAR system had an initial rollout in March, with the company sending emails of recommended cars to select dealers. In August, the system went live on the Adesa.com marketplace. KAR said it is working on a future service that would automatically purchase recommended cars on behalf of dealers and deliver them.
The company declined to specify how many of its recommendations turned into sales, but said the tool has resulted in tens of thousands of transactions since March.
KAR competitor Manheim, a subsidiary of Cox Automotive Inc., also began deploying various AI tools to dealers this year, including a recommendation engine slated to be generally available in February.
One of the early adopters of KAR’s recommendation engine was Virginia-based Pohanka Automotive Group, which has 17 dealerships, mainly in Maryland and Virginia. So far this year, through early December, Pohanka has purchased about 350 vehicles on Adesa.com, roughly half of which were a direct result of the AI-system’s recommendations. In the comparable period of 2018, Pohanka bought 110 vehicles through the site.
Scott Crabtree, president and partner at Pohanka, said the vehicles his company purchased via the recommendation engine sold in 37 days on average, compared with about 39 days for used cars obtained from other sources. That is notable, he said, because used cars depreciate while sitting on the lot, and Pohanka, like many dealers, finances some of the used-cars cars in its inventory and pays interest that accrues daily.
“What it means is that the theory is right, but we don’t have it fine-tuned yet,” Mr. Crabtree said. “But the fact that it’s better than just the normal process is encouraging.”
KAR said the recommendation engine is built on billions of data points from millions of retail and wholesale transactions, along with other market and economic data sources. This includes data from nearly 15 million transactions from both physical and online auctions that KAR recorded between 2014 and 2018.
Michael Ramsey, an automotive analyst and senior director at research and advisory firm Gartner Inc., said the industry’s embrace of AI demonstrates the growing appeal of technologies that can save costs and time.
“The key is how well does it work—if it works very well, then they’re going to get business,” he said. “Car dealers are already operating on extremely low margins, so if a platform like this can provide even a little bit of margin, it has a big value to them.”