Facial Recognition May Not Work For Security, But For CRM, ...

With all of the in-store changes being pushed by mobile—including the long-overdue predicted disappearance of the cashwrap—retailers are going to be moving payments in-aisle and pretty much anywhere in-store. Does facial recognition make sense? One vendor on Tuesday (April 23) tried, with a system that allows facial recognition to unlock a refrigerator, which determines which products are removed based on weight.

Although the current version of the system has serious technical limitations, the potential is probably greater for facial recognition than any other technology. Unlike PIN, cardswipes or any other form of biometrics, facial recognition has the promise to do far more than verify identity, including guessing at emotional state, gender, age and other attributes.

Selling suntan lotion? How about the ability to identify shoppers with deep tans? Want to be able to tailor suggestions made to first-time shoppers based on gender or age? With a thermal scan comes the ability to pitch hot cocoa to shoppers whose skin is especially cold or ice-cold lemonade to those whose skin is especially hot. (It could factor in the outside temperature so that it doesn't confuse someone who has been working in the sun with someone who is just running a fever. And if that shopper does have a fever, how about 50 cents off Tylenol?)

As a practical matter, facial biometrics has huge potential for CRM and marketing efforts, much more so than for payment authentication. ShelfX, the vendor behind Tuesday's introduction, is pushing its facial recognition package as a way for shoppers to charge their account when removing items from a locked refrigerator, instead of using the NFC card that shoppers are issued.

A minor restriction is that each store—not each chain—has its own customer database. That means that if a nine-store convenience chain installed this system, a customer would have to re-train the system each time a new store was visited. The much greater concern is the system's sensitivity. The system is not likely to recognize a shopper if there is even a slight change in appearance, said ShelfX CEO Ran Margalit. "Let's say you have on sunglasses and you didn't shave today," he said, citing two situations that could fool the system into not recognizing a shopper. "That's why we tell you to keep your card in your pocket." (Here's a rather low-cost demo of how it's supposed to work.)

If the system is that particular—and if the shopper has to bring the NFC card along anyway—it seems simpler to just use the card. Unless, of course, the shopper was given a monetary incentive to use the facial recognition system. It's no different than bribing shoppers to fill out surveys or to participate in other CRM programs.

The technology is rapidly approaching the point where these wacky ideas may become practical. In the same way that Google Voice is showing Siri a thing or two about getting voice recognition correct, experiments with facial recognition have shown sharp improvements in how usable data can be extracted. Unilever did a wonderful ice cream vending machine trial, where the machine only disposed ice cream to people who were smiling. (Frownie-puss? No ice cream for you.) Kraft ran a facial recognition supermarket trial, which used it to identify gender and issue different coupons based on that assumption. Pennsylvania tried using it for its state stores to determine whether a shopper was old enough to purchase wine. And a University of California study found shopping differences between sad and angry customers, giving a reason to have an automated means of differentiating between the two.

Wouldn't want to replace card-swipes with facial recognition, but in a year or two, having this specific form of biometrics as a CRM and marketing tool has serious possibilities.