Cars.com’s AI matches car buyers with new rides | Tech Industry
The car-buying business has changed a lot since Cars.com — one of the internet’s first classified car ads websites — debuted in 1998, but the Chicago-based automotive marketplace is getting with the times. Today it announced a new matchmaking experience that uses AI to help car buyers find the ride of their dreams.
“We’re treating people like human beings with distinct emotional nuances, not just site users, as we build a more relevant, personalized car shopping experience,” Cars.com chief product officer Tony Zolla said in a statement. “Early-stage car shoppers don’t know what they’re looking for. In fact, an overwhelming majority are undecided on make and model, yet nearly all online car search experiences force people to select make or model as the first step in their journey.”
The new search feature — built on the back of Cars.com’s 5 million car listings and 6 million editorial, user, and professional car and car dealership reviews — is aimed at the more than 70 percent of shoppers who haven’t yet decided on a make and model. Leveraging sentiment analysis and natural language processing, it offers a guided, algorithmic approach to recommendations, surfacing cars likely to appeal to individual sensibilities.
Here’s how it works: Rather than having users specify a car brand or series, Cars.com’s search tool instructs them to enter a price range, must-have features, and roughly a dozen “lifestyle preferences.” (Categories include “family taxi,” “daily commuter,” “fun and play companion,” or “workhorse.”) Then it presents them with a Tinder-like flow of 20 matches they can “like” or “dislike” at will.
Behind the scenes, machine learning algorithms take into account eight different categories and 28 subcategories of behavioral data, overruling shoppers’ stated preferences if it thinks they might consider a car similar to ones they’ve viewed or “liked” in the past. It also factors in car dealership data — if a matched model is only available from a dealer with a poor approval rating, Cars.com might not show it.
When a successful match has been made, the company funnels users to a car dealer they can contact to inquire about the car in question. (Cars.com takes a cut of sales leads.)
“Buying a car is normally a really confusing process,” Zolla said in a phone interview. “The more than 40 car manufacturers have 460 models and different trim levels at any given time — over 12,000, to be exact. That’s a number that most people can’t comprehend. Our approach simplifies things dramatically.”
The results so far have been promising. Since the new search tool entered beta, there’s been a 752 percent boost in user profile creation on Cars.com and a 225 percent increase in email leads, Zolla said. And more than 87 percent of visitors return to searches after they begin them.
“Technology companies have reset expectations,” Zolla told VentureBeat. “People think that the automotive industry hasn’t caught up to matchmakers like Netflix. We set out to change that perception.”
Cars.com is one of the largest car marketplaces on the web. It gets more than 400 million visits per year, and its unique monthly visitor count reached 25.6 million in June.