This kind of physical, outdoor advertising has been hemorrhaging dollars to online ads for decades for several reasons but a big one is online having a much better (but far from perfect) attribution story. Attribution is the big, thorny problem in _all_ of advertising e.g. how do you know if Customer A saw Ad B and that they then went on to buy Product C.
In my experience this type of tech has two goals:
The first is measuring distribution with aggregate data about how many people saw the ad and when, then combining this with ethnographic data about the location (e.g. what proportion of people in this area fall into income bracket Y). This looks like "1,293 people have seen your ad in the last 60 minutes and here some demographic information about them". This is done with video and gait tracking, or radar/lidar for measuring volumes and people who 'engage' with the ad by lookijng at it for >N $TIME.
This is all perfectly possible and scalable and is what ClearChannel are currently doing, the product trick is to make it 'real-time' so advertisers can adjust their buys and unit distribution more flexibly when combined with digital signage (e.g. I'm going to move our ad from this unit in this location to this other unit in this other location in 'real-time' as there's more high net worth people there today). This also nicely ups outdoor spend (people use it more, it appears more sticky) but they also charge more for these more 'dynamic' outdoor units. There's a bunch of seriously dubious data the the publishers use to justify this new way of buying outdoor.
The second (and WAY worse one) is doing direct attribution (e.g. Did consumer sKIFW3890 see ad sdoflij0934r and then did they buy product 029w3r29 from merchant 203ei2-309ir). This has been the holy grail of outdoor for a decade but no one's been able to make it work. ClearChannel very clearly WANT to do it, which is creepy as all hell, but in my reading between the lines (and working in the industry a little) I don't _think_ this is what their Radar solution delivers, at the moment. I could be wrong, though.