"We go down to the feature level when we analyze these cars. We identify local competition," Chad Bockius, Vast's chief product officer, told Automotive News. "We have an algorithm that allows us to identify vehicles that are most similar. We analyze the competition the same way we analyze your vehicle. We look at recent sales history. We try to get a sense of how quickly those cars are moving compared to the rest of the market. Looking at the features of those cars, which elements are driving them to move more quickly or slowly?"
Bockius compared CarStory Insights to IBM's Watson, an artificial intelligence platform that once competed on the TV show "Jeopardy." For each question, Watson generated three answers with varying degrees of confidence — and even answered incorrectly sometimes.
Bockius said Insights takes a similar approach. The platform will generate predictions for vehicles with considerable data available in certain markets. For rarer vehicles that are sparse in a particular market, Insights will remain silent because the amount of data is insufficient to make a strong prediction.
The app will judge how a price change and even color will affect sales.
"We know what features drive churn. You might have a vehicle not moving very quickly and part of the reason is you only have five or 10 features that are affecting churn," Bockius said. "Next thing we look at is color. You might have a blue Jeep Grand Cherokee when people want white, black and red. That's all based on our analytics."
Harry Haber, used-car manager for Capistrano Mazda and Capistrano Volkswagen in Southern California, uses Insights as a supplement to the vAuto inventory management solution.
Sometimes the tool will recommend a price that Haber considers too low, but that isn't the app's fault. In these cases, Haber said, he realizes that the store may not have had the right game plan for the vehicle.