CarStory launched a product Tuesday that uses artificial intelligence and machine-learning, along with predictive analytics, to gauge when a car will sell and project its sale price.
CarStory Insights is designed to predict incoming trends and give dealers guidance based on that predictive insight. Through analysis of a dealer’s lot, it offers alerts, sales predictions and pricing recommendations.
“It all started from conversations that we were having in the market with dealers. We really wanted to understand what problems weren’t being solved. What were the things that dealers wished they would know, but they couldn’t get today?” said Chad Bockius, who created CarStory, in a phone interview Monday.
“And the common thing we heard was that dealers have access to more information than they ever had, but they still don’t have answers,” Bockius said.
“And we got to the point with all the data that we’ve collected, all the analytics that we’ve been working on, and now where, just the state of artificial intelligence and machine learning, that for the first time, we can actually give dealers answers, not just give them more information.”
Use of artificial intelligence
When you hear the word “artificial intelligence,” you might think Will Smith or Tom Cruise movies full of robots and dystopian futures.
Chad Bockius, CarStory
But in the case of CarStory Insights, think “Watson” — as in, the computer platform from IBM that competed on “Jeopardy!” in 2011.
Bockius said that with this new platform, CarStory is “doing the same thing that Watson did.”
What Watson did on the TV show was to generate three answers whenever there was a question. Each answer had a different level of confidence.
“And if the level of confidence in the first answer was high enough, Watson would ring in and deliver that answer. It didn’t’ always get them right, obviously. There was never an answer where Watson was 100-percent confident, but a lot of them were very high,” Bockius said.
“We use AI in the same way to actually generate those sales predictions as well as price predictions,” he said. “So for us, we need to look at all that data that we were able to pull together, and then say, ‘When does the machine believe this vehicle will sell?’
“And the machine is either going to have a very strong belief that it’s correct, or it will be weak. And obviously we’re reporting on the safest cases where the probably is strong, where the confidence is strong.”
The company also uses AI to run what-if scenarios to determine, for instance, how much the car’s price would have to be reduced to make it sell faster.
Bockius also noted: “At the end of the day, when you talk about things like machine-learning and artificial intelligence, this really gets down to the types of algorithms that we’re using under the covers.”
Machine learning, he said, is a crucial piece to the AI stack. Through machine learning, the company can feed the computers wide swaths of data and then use the algorithms to pick out “interesting points,” Bockius said.
“So for example, one of the things that we discovered is really important is to figure out which features on a car affect whether or not it will sell quickly. We call those ‘turn-drivers’ … and that was one of the ways that we used machine learning to feed into these predictions about when a vehicle will sell,” he said.
“Because obviously, the more features that affect turn a car has, the more likely it is to sell quickly,” Bockius said.
“So, then you get to a point where you aggregate all of this data to essentially start making predictions.”
'Predict the future'
With the CarStory Market Reports product, the company has, for a while, been examining what demand is expected to be for specific vehicles. And that can be impacted by factors like seasonality, gas prices and more.
“And we can start to see behavior changes online that help our predictive model” determine what may happen, he said.
“Now that technology is at a point where it’s cheaper and easier to do machine learning and AI, and we actually have the data to train the computers, it would be silly to continue to rely on historical averages if you can predict the future,” Bockius said.
“And I think that’s what we’re going to see, not only in the space we’re working in, but I think in every space across automotive, you’re going to see this being applied, whether it’s in customer service or in marketing or in sales or in sourcing inventory,” he said.
“There’s simply no area that’s going to be safe from disruption when you have this ability, because it’s just a better way of doing it. No one would sit here and say, ‘I don’t need the answer, I just need more information.’”
5 features of CarStory Insights
In a news release, the company lists these as key features of the product:
- Turn Drivers: Highlights the vehicle features that help cars sell in a local market and compares those to every vehicle on a dealer’s lot.
- Smart Pricing: Provides pricing recommendations based on how each price will change the days to sell.
- Market Rank: Provides a detailed comparison how each vehicle stacks up against the local competition.
- Sales Alerts: Get notified via phone or email whenever Insights detects an issue in the market or on your lot.
- Instant Share: Notifications can be easily shared with members of a dealer team.
When asked specifically about the sales alerts function and the issues in the market that might arise, Bockius said the alerts boil down to a slow-seller alert and a fast-seller alert.
“Now, there can be a lot of things that cause those alerts to be triggered. For example, you might have Hertz dropping 200 Honda Accords on the market, and all of a sudden, that dramatically changes when those Accords on your lot are going to sell,” he said.
“Nothing changed with your vehicle obviously, but the market changed. Alternatively, we could see consumer demand going up or down, and again, that could have a dramatic effect on when your vehicle is going to sell. So what we do is we take all that complexity, we look at all the signals, make sense of them all and then simply alert the dealer to a problem.”
Like, for instance, your vehicle might take longer than expected to sell.
Then it gives guidance to the dealer on what they would have to do regarding price for that car to move faster. And the same goes with cars turning too quickly.
Currently there are 1,500 dealers on waitlist for CarStory Insights, Bockius said.
It will be rolled out first to CarStory Market Report Customers. The company will be activating the product for dealers who were early beta participants, then opening it to the wait list. The goal is to have it open to everyone by the end of the year, Bockius said.