This marriage could be a very low-cost proposition, with the app asking for the cart's number or, better yet, the cart presenting a QR or barcode that the app can scan. The app—with its capabilities and its CRM history—could use the cart equipped with a simple radio beacon for navigation purposes.
This approach would certainly have its limits, not the least of which is that many chains don't even use carts. It is also not practical in malls. And, of course, plenty of customers don't use shopping carts—or at least don't use them all the time.
But this relatively dumb cart approach is a lot cheaper and more efficient than smart carts, which have been slow to catch on and to do much ROI "make CFOs happy" magic. The Microsoft-fueled smart cart, demoed in Whole Foods while that chain stressed it would never use it, is a wonderful example of something that is a lot less practical. For instance, remember that shopping carts don't shut down in sunlight like Dracula. (That sunlight-shutdown feature alone makes the Whole Foods shopping cart story worth its two-minute read.)
The biggest advantage of this type of approach, though, is its near-term tactical advantage. Tracking mobile devices directly is superior in the long term; it's also strategic. The advantage of mobile-only is the consistency of information as the customer moves from store to store, especially over time.
The problem, though, is that retailers need approaches they can fund and deploy in a few months, not years from now. The cart concept could be useful in getting something operational soon.
Let's not forget that, as the commenter pointed out, it's not going to be easy to have a consistent method for tracking Apple and Android devices, let alone the variations among many of the Android manufacturers. For that matter, even the Apple hardware side will have its variances from newer models to older units. Then there are BlackBerries and the legions of other phones coming.
It's already clear that no single physical-tracking technology has solved the problem yet, which means for now tracking will have to depend on Wi-Fi triangulation and audio triangulation and route tracking using direction changes and accelerometers, for example. But retailers have no control of which of those features a customer's smartphone has. On the other hand, a cart-tracking system is completely under the store's control, so the chain can decide what combination of in-store location approaches works best and cheapest.
Once shoppers can be tracked through mobile apps and something else such as a cart or basket, the potential for changing the in-store experience is huge. As Walmart is working through its in-aisle mobile checkout experiment, this means texting promotions to customers within seconds of them taking an item off the shelf, in plenty of time to change their behavior.
It enables the tracking of what customers pick up, put in their carts and then put back. It can note exactly where shoppers were when they changed their mind (what was the digital signage saying at that moment?) and how long it took.
It's simply that the elegant method of spying on phone signals may not work very well in the near term. So if stamping a shopping cart gets us these advantages now—while we're working through the bigger mobile issues—maybe that's worth considering.