What You Can Learn About Online Shoppers by Watching Them
People who shop between midnight and 1 am are twice as likely to be fraudsters, according to aggregate retailer data from Sift Science.
(No, officer, don’t profile me — I swear I was just doing some relaxing retail therapy before drifting off to sleep.)
While you are shopping online from the seeming privacy of your home, many retailers are keeping track of your every move. And in aggregate, data about online shoppers show some interesting trends and habits that startups are emerging to track on behalf of retailers.
This morning, I covered the launch of Sift Science, a brainy startup from ex-Googlers that’s applying machine learning to detect fraud patterns in online retail.
Sift Science co-founder Brandon Ballinger told me that during beta testing with services like Airbnb and Uber, his company had observed a million different signals that flag any one buyer as a potential fraudster.
For instance, beware of people who try to buy something with a Yahoo email account — they’re twice as likely as the norm to be fraudsters — but users of AOL and Outlook.com email domains are much more likely to be safe.
In a similar vein, I recently met the founder of Commerce Sciences, Aviv Revach, whose company tries to apply behavioral science to increase sales on e-commerce sites. The company supplies an overlaid bar for online stores that is personalized based on observations about an individual user’s visit — for instance, browser type, mouse movement patterns and the specificity of a search term.
“There’s so much that science knows about us that business doesn’t use,” Revach said. For example, he pointed to theories of “hedonic” versus “utilitarian” tendencies in consumer consumption, where it might be more effective to have an entirely different shopping experience for something fun versus something functional.
More specifically, Commerce Sciences can recognize signals that a certain user at a certain time is more driven by sales, or more influenced by his or her friend’s opinions. Then it reformats the bar to push that angle. The goal is to grab the more than 90 percent of people who enter an online store and exit without buying.
I asked each company for fun, odd and interesting examples of things they’d learned from observing shopper behavior. Be warned that both companies have only just become available in the past year, so their data represents a mere slice of online retail. Here are a few from Sift Science:
- People who type their last name in all caps are 5.6 times more likely to be fraudsters.
- People who don’t use Facebook to log in are four times more likely to be fraudsters.
- People who log in from any one of eight particular countries are 219 times more likely to be fraudsters.
By the way, Ballinger noted that stores only need a user’s credit card number and expiration date to process a payment, but they often ask for more information for their own purposes or to try to combat fraud. “There’s a trade-off between fraud and friction,” he noted, which is a nice way to put it.
Meanwhile, here are some findings from the early days of Commerce Sciences, from testing on online stores that sell apparel, flowers, music and other items:
- Social proof — showing what your friends liked and bought — appears to be much more effective at night. In tests, it resulted in 49 percent more sales in the evening hours, and it didn’t help daytime sales at all.
- People who are referred to an online store from Google are especially susceptible to 10-percent-off discount coupons, spending twice the average amount on some sites.
- Adding the word “free” — “You have won a coupon” versus “You have won a free coupon” — increases sales by 15 percent.