Skip to Content

Watching Social Networks for Clues about Promotions

A company founded by database engineers helps retailers figure out the likeliest way to get people to buy things.

Before people began making their lives public on social networks, retailers had to figure out their likes more indirectly. Companies like Oracle and Siebel provided huge database programs that analyzed individual sales and consumer demographics in search of patterns that might lead to more sales—say, by targeting groups of customers who might be likely to respond to special offers. But the retailers couldn’t directly observe the connections between individual consumers, or watch them chat online with each other about products.

Retail investor: Computer scientist Vivek Subramanian cofounded the Silicon Valley company CalmSea, which helps retailers use social media to target promotions.

CalmSea, a company founded in 2009 by database industry veterans, adds consumers’ social-network activity to the mix of what retailers can analyze. Moreover, the company takes advantage of cheap large-scale database computing to crunch numbers nearly in real time, and to experiment on the fly with new analysis methods that would have required serious database retooling in the past. Clients include the sneaker company Puma and Tobi.com, an online clothing retailer.

“Retailing is changing,” says CalmSea’s vice president of products, Vivek Subramanian, himself a former engineer for Siebel’s analytics software. “We used to look only at data inside the enterprise. Today, there is a lot more data coming from outside the enterprise.” The primary source is Facebook, where CalmSea clients publish apps branded under their own names. An app for Tobi.com, for example, may offer an extra 10 percent off sale items; when someone uses the app, the client can see which Facebook pages for brands, products, or retailers the customer likes, shares, or comments on.

By giving each customer a unique promotional code, CalmSea can track who buys what. It can also see who a customer’s friends are, making it possible to identify groups of like-minded shoppers who may not even realize what tastes they have in common. The retail industry “used to look at past sales data on who bought Ralph Lauren jackets in April,” Subramanian says. “Now we combine that with the social graph of current activity. What are the most popular brands followed in their social network? Maybe it’s Calvin Klein instead.”

CalmSea calculates individual customers’ brand and product affinities, as well as their “promotion elasticity”—that is, how big a difference a discount makes in their propensity to buy a promoted product. Some customers won’t buy without a markdown. But even in the era of Groupon super-discounts, sellers are wary of giving everyone a price cut that might not be necessary. Price optimization is one of the reasons retailers spend big money on databases.

CalmSea’s technology analyzes both traditional customer histories and social-network behavior to help retailers and brand managers identify the right offer on the right product for the right customer, via the right marketing channel. CalmSea clients use a dashboard that groups customers by their likes, behaviors, and friend networks. Then the clients can create promotions—discounts, group buys, or perhaps a sweepstakes in which customers get an extra entry every time they share the promo. The dashboard also makes it easier to target promotions to specific groups of customers defined by gender, age, and other characteristics. The retailers then publish the promotions—essentially, new apps—on their websites, on Facebook, and on Twitter.

When potential buyers accept the new promotion, granting it access to some of their Facebook data, CalmSea then tracks the response. How many people clicked on it? How many liked it, or shared it on Facebook? How many bought something, and what did they buy? How much did they spend? CalmSea merges Facebook behavior with additional retailer data, such as information on product returns, to create an interactive chart showing how well the promotion worked at each step of the process.

Tobi.com’s vice president of marketing, Jennifer Song, says she can pinpoint the benefits. “One specific promotion that we had huge success with was a discount sent to our existing e-mail base, to encourage them to also become fans on Facebook,” she says. “We saw an immediate 500 percent lift in revenue from Facebook. Seventy percent of the participants actually went on to convert into sales.” More important, these customers can now be targeted in future campaigns by taking into account how they reacted to the offer.

CalmSea’s tools are built atop two cost-saving, easy-to-scale database technologies: Hadoop, an open-source system capable of analyzing petabytes of data across thousands of servers, and Amazon’s Elastic Compute Cloud (EC2), a service that lets clients rent number-crunching capacity by the hour. The flexible Hadoop framework allows CalmSea’s technical team to test new modeling methods on the fly, without reconfiguring the database first. And by doing most of their processing at night, during EC2’s off hours, they keep operating costs down. Five years ago, Subramanian says, “we would have had huge capital expenses for servers just to do testing. Those machines cost a million dollars, and we used to install them with a crane. Paying by the hour on EC2, a machine costs us a few hundred bucks a year.”

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.