well, to continue playing devil's advocate, we've had hundreds of years of advertising (newspapers, pamphlets, and posters/signs before that) without any kind of systematic data collection, so it's not necessarily an intrinsic characteristic of the advertising business to collect data.
in an alternate universe, if banks didn't have such a stodgy and stifling rent-seeking business on the movement of money (transaction fees on everything), microtransactions might be the preferred business model of many online businesses that are now dependent on ads.
Anticipating needs of person X, (and serve person X an ad), without any foreknowledge of person X.
a made-up example might be a customer at best buy spending time in both the tv and pc departments and then having an ad pop up on a nearby screen for an intel NUC and/or apple tv (making the assumption you want a tv-connected computer). the ad didn't need to know any personal characteristics about you other than limited, directly observable behavior in the store.
i'm sure people who focus on this every day (i.e., not me) can provide better examples.
You'd have to recognize the face and track it prior to displaying the ad.
if you went that route, you could get away with simpler person recognition (height, width, clothing color, accessories, etc.). or you could just track phones and other devices via wifi/bluetooth (and presumably uwb in the near future).
but you could also just have salespeople on the floor doing the recognition & ad placement part of it, for a lower-tech solution.
i'd actually bet trained salespeople would have better hit rates over algorithmic solutions, where access to large amounts of dislocated data is the main advantage (at the cost of our privacy).
fwiw, i generally don't want to be subjected to ads either, particularly for the privacy-invading aspect, but also the time- and attention-wasting aspect too. most ads are absolutely horribly conceived and targeted, especially spray-and-pray email spam.