Those sweaters on sale at your favorite department store could be in the process of being optimized by an M.I.T. spinoff.

Boston-based Celect, founded in 2013 by two M.I.T. professors, is today announcing it has scored $5 million to help retailers select and distribute their products in ways that reflect buying patterns.

The company’s core product is the Choice Engine, based on customer choice modeling.

Choice modeling, CEO John Andrews told me via email, “is the technique of using customer and transaction data to model out how a customer is likely to react to and choose from a set of options.”

Cofounder, CTO, and M.I.T. professor Vivek Farias added that the company’s platform is complex, but the core concept “is very simple.” The other cofounder is chief scientist and M.I.T. associate professor Devavrat Shah.

“When a customer buys a product,” Farias emailed, “you now understand that customer’s ‘selection.'”

“But what if you also knew not just what a customer bought, but also what was available to them when they made their selection? Said another way, what if you also knew what they didn’t buy?”

That would mean, he said, you now know their preference, not just their selection.

A Celect screen, indicating product optimization for a retailer.

Above: A Celect screen, indicating product optimization for a retailer.

Image Credit: Celect

For large retailers, there are a lot of choice possibilities. A retailer with 500 stores, for instance, could have five million products to offer online and 150,000 to 200,000 for any one store.

The platform uses the buying patterns to construct recommendations that are delivered through a Web interface. APIs deliver the recommendations to a website or other enterprise systems.

But the recommendations, Andrews pointed out, are not visual merchandising suggestions about which sweater should be put where on which store shelf.

Instead, it recommends things like which products the store should stock, how many of each, how they should be allocated to each store, and what’s the best strategy for marking down their prices.

Online, there are real-time recommendations on which products a specific user is most likely to buy — at each click as she navigates the site.

The result, according to Celect, can be an upwards of 20 percent lift in online sales and a seven percent increase for in-store revenue.

The data crunched to find these patterns includes transaction logs from stores and websites, inventory info on what was available when a selection was made, product catalog detail, and anonymous customer information on who bought what over what period of time.


VB just released Marketing automation: How to make the right buying decision (the first time)
$299 on VB Insight, or free with your martech subscription


This series A round, which adds to a previously raised $500,000, was led by August Capital with participation from Activant Capital Group. The money will be used to grow the team in engineering, sales, marketing, and customer success.

As for competitors, Andrews says they aren’t the human retailers.

“Retail, and specifically the merchandising and planning process, is very much an art form,” he acknowledged. Retailers are very skilled at understanding their products and customers, he added, but Big Data can supplement a merchant’s gut intuition.

It’s not replacing the art part, he said, but it’s “bringing science to the art of retail,” which also happens to be the startup’s tagline.

As for computer-based choice modeling, he noted that it’s been around for a while.

“All the big software players have solutions here,” he said, pointing to IBM, SAP, JDA Software Group, and particularly Oracle. But those are all “focused on Assortment Management, or in-store POS [point-of-sale] and online commerce platforms.”

“Our ability to execute at very high scale with very limited data on each customer is where we have innovated.”