QR Codes Are A Terrible Idea. Why Is Image Recognition Even Worse?

QR codes are ugly. They're intrusive. Most designers hate them because there's no way to make them look any less like the brick-full-of-blocks they are, especially when they've been slapped next to a great-looking retail marketing image. That's why the idea of leaving out the QR code entirely and just getting a mobile phone to react to the image itself is so appealing. It looks so much better that it's easy to forget why it's a bad idea: That ugly, intrusive QR code screams "Point your camera at me!" An ordinary image doesn't.

As a result, if potential customers know what they're supposed to do with a QR code, they can easily do it. But how are they supposed to know that there's any special significance to the image in an ad or porter or brochure?

This comes up because of a story we saw about LTU Technologies, which claims its image-matching tech is so good that just pointing a a smartphone's camera at an appropriate image can trigger the same kind of response—a web page, a coupon, additional details—as a QR code. We'll take the company at its word for that (though of course if you're thinking about using it, you'd be crazy not to test it thoroughly).

Unlike digital watermarking, this is pure image recognition. You don't have to load the image up with too-subtle-for-the-human-eye markers, you just upload the image to LTU along with indications of which parts of the image are most recognizable. It does the matching and sends the customer whatever information or redirect you've specified.

It's a clever idea, and it seems like it should work perfectly once it's everywhere. The customer points her phone at a logo and gets a website, at a magazine ad and gets a coupon, at a billboard and gets directions to the nearest story.

But it's not everywhere. And the vast majority of logos, ads and billboards aren't waiting for customers to point a phone at them. Point the phone and nothing will happen. After a while—a very short while for most customers—the charm will wear off.

It's hard enough to get consumers to point their phones at QR codes, where they can be sure something is encoded in the brick-full-of-blocks (even if it's something the consumer really doesn't want). They're slowly being trained by a few specific applications such as Peapod's virtual grocery stores and the occasional giant billboard, but it's QR's use is nowhere near what its proponents expected.

And that's with a technology that trumpets the fact that it's there. Pure-play image recognition, where the few images that can be recognized are vastly outnumbered by the great many that give no response, hasn't got a chance—at least out on the street.

Then there are the technical issues. A QR code actually contains information that can be decoded without a network connection or a vendor on the other end. In a cell dead spot with no Wi-Fi? Customers can point cameras all they want, and nothing will happen. (And there are a surprising number of dead spots, even in urban areas.) Nothing discourages the use of technology like that tech mysteriously stopping and then starting again.

That said, once you get customers in off the street and into a situation you control, this might work brilliantly. In-store, with customer-accessible Wi-Fi and images you've already uploaded to a vendor, customers can point at lots of images and get a reaction. You could even pitch it as a reason to come into the store. Prepping all those images is plenty of work, but so is adding QR codes to everything. And trying out image recognition might distract some customers from actually shopping, but that's a risk of any in-store technology.

Of course, you could get around the most-images-don't-work problem by putting some kind of logo near images that trigger recognition—something that flags those images as special. Something like, say, an ugly, intrusive brick-full-of-blocks?