Desired Features by Shoppers

Sleuth-ing Out Hidden Hits: Celect Tames Retail's Unruly Inventory

Ever wondered how big-name stores decide what items to stock? The answer doesn't lie in a crystal ball, but instead in some smart data analytics, thanks to MIT spinoff Celect. This intrepid company is revolutionizing retail's ancient practices with a slick software that deciphers purchase patterns, helping stores optimize their shelves.

Co-founded by MIT professors Vivek Farias and Devavrat Shah, Celect concocts digital wizardry that, armed with a store's sales and inventory data, determines which products are tailored to the tastes of local customers. The brains behind this innovative tech also happen to be the architects of Netflix's recommendation engine and the predictor of trending Twitter topics. They've taken their gift for unraveling hidden connections between items and set their sights on retail.

Celect's software scrutinizes the items parked next to each other within a store, examining sales records to divine which products are more likely to catch customers' eyes. When analyzed at scale – think thousands or millions of item comparisons – this method sheds light on the purchasing inclinations of the broader customer base.

"We're shrinking the world into a bag of comparisons and then bingo! Out pops our customer-choice engine," Shah, an associate professor of electrical engineering and computer science, and Celect's chief science officer, reveals.

Once retailers plug in budget, shipping, and other parameters, the software's user-friendly interface calculates the optimal arrangement of items to stock, taking into account the overall profit. Unforeseen treasures, such as church hats, may find their way onto shelves, pushing aside more mainstream, but less profitable goods.

"It's no mean feat behind the scenes, but what we're doing in essence is simple: we're tracking who wants to buy what, and ensuring the right product finds its way to the right person," says Farias, Celect's chief technology officer and a professor at the MIT Sloan School of Management.

Insights Over the Counter

Celect, which saw the light of day in 2012, has attracted a corps of big-name retail clients, some with hundreds of stores scattered across the nation. Last month, the company pocketed $5 million in venture capital, indicating retail bigwigs are betting big on their predictive prowess.

The software has yielded surprising insights into consumer tastes. In one case, a Midwest retail chain discovered that it ought to peddle church hats. analysts noticed a link between the purchase of soccer shorts and certain watches in specific stores and fancy hats for church. The headache of oversupply gave way to the heavenly profit of speedily selling church hats. The retail chain is now considering incorporating Celect's technology across its 270 stores nationwide.

Comparisons between the sales data of various stores have given birth to new discoveries. For instance, by comparing sales patterns in two stores, Celect can deduce that customers with shared preferences are likely to seek out an item in one store that wasn't available in the other.

In another instance, a retailer presumed that an older piece of clothing would sell less than newer items. But Celect's analytics unmasked a surprising truth: the item's popularity hinged entirely on its color.

"These kinds of nuggets just jump out of the data," Farias notes.

A Universe of Choices

Tracing its origins to a decade ago, when Farias and Shah aimed to enhance recommendation engines for Pandora, Netflix, and other online services, Celect's technology has been steadily tracking our purchasing missteps and decode our tastes as a species.

The key to Celect's technology lies in its ability to analyze paired comparisons, such as comparing movies and products. In a series of papers from 2008 to 2011, Farias and Shah presented an algorithm that used this choice model to better predict people's preferences. In 2012, they deployed the algorithm to predict trending Twitter topics with uncanny accuracy.

A case study published in Management Science in 2011 demonstrated the algorithm's potential for real-world application, by predicting car-buying preferences more accurately than other choice models. The algorithm analyzed transaction data over 16 months on a specific SUV offered at dealerships across the Midwest, predicting the "conversion rate" of turning an arriving customer into a buying customer.

It came as no surprise when Farias and Shah saw retail as fertile ground for their innovations. "We realized this isn't an issue limited to cars; it's fundamental to all choices," Farias remarks.

To commercialize their research, the researchers had to simplify their massive computations to make it feasible for everyday computers. They devised a method that took hundreds of thousands of possibilities into account, boiling it down to a manageable subset. "We managed to make it scalable," Shah adds.

As Celect embarks on its journey of growth, it has attracted top talent from MIT, other institutions, and technology firms, while building a team of retail industry veterans. Farias and Shah are quick to acknowledge the crucial role played by MIT's entrepreneurial ecosystem, including support from MIT Sloan and the Department of Electrical Engineering and Computer Science, in spinning their complex technology out of the lab.

Farias, with his roots steeped in academia, and Shah, a first-time entrepreneur, are eager to see their research make a larger impact on the business world. "It's a thrilling prospect to take science from the lab and see it make a tangible difference in the market," Shah muses.

A world where shelves brim with what we truly want, where the stockroom no longer holds any secrets – such a world might be just around the corner, thanks to Celect.

The MIT spinoff Celect, co-founded by professors Vivek Farias and Devavrat Shah, is revolutionizing retail with a software that deciphers purchase patterns, helping stores optimize their inventory.
This software scrutinizes the items placed next to each other in a store, examining sales records to figure out which products are more likely to catch customers' eyes.
Once retailers input parameters such as budget, shipping, and others, the software calculates the optimal arrangement of items to stock, taking into account the overall profit.
Celect's technology has attracted big-name retail clients, some with hundreds of stores, indicating retail bigwigs are betting big on their predictive prowess.
The software has yielded surprising insights into consumer tastes, such as discovering that a Midwest retail chain should sell church hats.
Comparisons between sales data of various stores have led to new discoveries, like deducing that customers with shared preferences are likely to seek out an item in one store that wasn't available in the other.
Farias and Shah, who aimed to enhance recommendation engines for online services a decade ago, have been steadily tracking our purchasing missteps and decoding our tastes.
Celect, with its team of retail industry veterans and top talent from MIT and technology firms, is eager to see their research make a larger impact on the business world, particularly in data-and-cloud-computing technology, finance, retail, and other industries.