As far as e-commerce has come, it still remains in its infancy. A glance in any spam folder is proof positive that online retailers haven't yet refined their customer tracking. To wit: My spam box currently features Petco.com  advertisements for kitty litter (I'm a dog person), a Staples.com  ad for Windows software (I'm a Mac girl), and four ads for Viagra (enough said).
But the emails from Barneys.com  are different. Barneys knows that I like jewelry and yoga. My most recent Barneys email read, "Love it! Jennifer Meyer Ohm Necklace." I do love it.
In the past eight months, Barneys' relationship with Sheldon Gilbert, a genetics scientist turned software impresario, has given the retailer the ability to precisely target customers in its email campaigns. Gilbert's company, Proclivity, sorts through the data left by millions of anonymous people clicking around Barneys' Web site, and predicts who's likely to buy which products, when, and at what price.
"A lot of companies throw this data out, or only use 1% of it," says Gilbert, 32, a suave St. Lucia native who studied molecular biochemistry and biophysics at Yale and spent two years doing genetics research at Cornell Medical College and the Rockefeller University Laboratory for Molecular Genetics and Informatics before segueing into the private sector. Stints at a company that built Web sites for J.Crew, Best Buy, and Martha Stewart, and later working for discount apparel retailer Bluefly.com , made him realize the wasted potential in not mining site data. He deferred grad school at NYU in genomics to pursue what became Proclivity. "Scientists understand how complex systems work," he says. "I'm a pattern hunter, so I created a system that was looking for patterns and was adaptive and self-learning."
The impact on Barneys, Proclivity's first customer, has been significant. "We used to spend $90,000 on a full-page ad in The New York Times," says Heather Kaminetsky, Barneys' director of Internet marketing. "Then with the Web site, we would send an email about, say, Lanvin handbags to 100,000 customers. But 90,000 of them probably didn't even know what a Lanvin handbag was." Today, only those people Barneys has identified as handbag fanatics get an email, and Barneys has seen up to a tenfold rise in response rates.
Kaminetsky can target customers based on their overall habits, such as "fashionistas" who buy risky new designer products, "bottom feeders" who always buy sale items, or cosmetics zealots. "We know when you're gonna run out of shampoo, so we might as well send you an email," she says. So she does. Rather than feeling spied on, customers are thrilled, because the message is relevant. Who wouldn't want a reminder that it's time to get more shampoo?
Barneys was skeptical about Proclivity when Gilbert approached the company in the summer of 2007. "When a vendor comes to you using all those words -- 'integration,' 'low-tech,' 'turnkey solution' -- it's scary," says Larry Promisel, VP of e-commerce at Barneys. He is bombarded by sales pitches from similar companies, many of which track basic customer patterns. But Proclivity's software brought all of the information together, and took the further step of predicting customer behavior. "We tried to trip up their system," Promisel says. "Sheldon would suggest which designers would be big for us, and we'd separately make our own judgment based on the store. And he was right." Once the firm was hired, Proclivity spent two months collecting data -- letting the computer model learn (the early predictions were sometimes less than accurate) -- and another four months testing. "We had to teach them a bit about luxury retail. They butchered the names of our designers," Promisel chuckles, citing Dolce & Gabbana and Manolo Blahnik as victims.
The Proclivity analysis has born unexpected results. Barneys now knows what time people like to buy (lunchtime) and which products are likely to sell well in pairs (handbags and wallets). And based on Proclivity's feedback, Barneys also redesigned its e-campaign formats. Unlike the kind that cram in 5 to 12 products ("Spring Tie Sale!"), a typical Barneys email sticks with one product from one designer: "Manolo Blahnik: Pretty in Pink" features an image of a sparkling pink high heel on a pedestal. That's it.
Barneys is considering expanding Proclivity to its stores -- Gilbert's program tracks products as well as customers -- to marry its in-store and online marketing efforts. And with such an accurate predicting system, Barneys is now exploring the limit of how many emails are too many. Customers, it turns out, are much more tolerant of e-advertising when the products being hawked actually appeal to them. "We try to limit it so every person doesn't get more than three to five emails a week," says Kaminetsky. "But around holiday season, it's a free for all."