The Internet offers huge opportunities for buyers and sellers of used cars, but one truism still holds: Nobody wants to buy a used car sight unseen. That fear of paying for a lemon online has helped keep used-car dealerships in business, despite consumers’ distaste for the hard sell long associated with such venues.

But Carvana, a Phoenix-based online car marketplace launched in 2013, saw in Big Data multiple opportunities to better consumers’ experience and its business. Among the company’s 50 staff members are some you would never find selling cars: It employs five data experts, including a former Wharton professor who oversees Carvana’s analytics.

Well before Carvana launched, the company contacted Kaggle, an online community of data scientists who compete and collaborate to solve Big Data challenges submitted by companies such as Merck and Facebook. Carvana wanted a better way to predict if cars purchased at auction would be lemons-“kicks,” in used-car parlance-and it anted up $10,000 in prizes for the best solutions.

A system emerged that kicked the kicks and let Carvana make shrewder bids at auctions. “We wanted to determine which cars up for auction didn’t meet our quality standards,” says co-founder Ernie Garcia. Thanks to the winners of Carvana’s Kaggle competition as well as data gleaned from other analytics about regional customer preferences and model availabilities, Carvana found it could sidestep lemons and buy better cars “for $500 below what similar cars would sell for.” That edge is helping Carvana meet its goal of offering its customers an average $1,500 discount from market prices.

Carvana also worked with its data-expert staff members to mine customer data and reduce the risk in its financing business. Although many car dealerships simply look up a buyer’s credit score, Carvana scans hundreds of variables across several databases-including full credit reports from multiple companies and searches of the LexisNexis legal and news databases-to predict the likelihood of defaults, better tailor interest rates for individual customers, and weed out suspect buyers.

The result, according to co-founder Ryan Keeton, is “meaningfully” fewer defaults and not one car stolen through fraud. (Citing competitive concerns, Carvana executives did not share specifics concerning default rates or total vehicle sales.) Online used-car marketplaces still haven’t replaced the banner-strewn used-car lots of yore, but Big Data is helping one such venture find its footing. As it also helps a constellation of other smaller businesses, be they old or new.