MIT Startup Celect Gets $5M to Apply Predictive Analytics to Retail
By now you’re used to the idea that companies are trying to predict your every move when you shop online. But how about in the real world?
Boston-based startup Celect has a new approach to understanding and predicting customer behavior in brick-and-mortar stores. Now the company has raised a $5 million Series A round led by August Capital, a Bay Area venture firm, to advance its ideas. Activant Capital Group also participated in the deal.
Celect is trying to solve a fundamental problem: how should retailers stock their shelves to optimize customer purchases? That means putting the right product assortments in the right places, at the right prices, with the ultimate goal of moving more inventory.
The company was founded by MIT professors Vivek Farias and Devavrat Shah. Farias specializes in operations and revenue management, while Shah is a network theorist who researches the infrastructure behind social networks and data flow. Celect is led by CEO John Andrews, a veteran of Oracle and Endeca.
There’s no easy answer to the product arrangement question, but plenty of ways to approach it using data. Farias and Shah have studied “choice modeling,” the science of how consumers make decisions. In their research, they have used machine learning and other techniques to predict revenues based on specific assortments of product choices (in the automobile industry, for example).
Now their company is applying those results to retail stores, looking at things like purchasing data and the context around it—and finding patterns in who buys (and doesn’t buy) which products, at which prices, and in which product groupings.
Celect’s technology has been in pilot tests with 10 Bon-Ton department stores around the country, and the company says the software “shows positive increases in in-store revenue over the control group.”
It’s still very early, but the big goal is to boost sales and, eventually, create a more personalized shopping experience for consumers. That means being able to predict what they want—and, better yet, what they don’t even know they want.