So far as I can tell a large fraction of adverts "targeting" me don't understand time's arrow.
I bought some very expensive high heels as a treat for myself. They don't make them in my size, I bought them because they're beautiful and put them on display.
In the following weeks I received lots of advertising for those exact shoes. Not for high heels generally, not for nice frocks, and certainly not for glass display shelves, just endless adverts for the same exact pair of shoes.
Somewhere a "digital advertising executive" is telling their client that these adverts were very highly correlated with sales of the shoes. Of course, they'll disclaim, correlation isn't causation. But it is - that's their dirty secret, just the reverse of the causation pretended to. Advertise to people who are already buying the product, then say "See advertising works".
The shoes are the most blatant example because they're so noticeable (I did not spend that much on shoes which are merely a pretty red colour or covered in sparkles) but I even saw this for mundane things like a Seagate hard disk. I purchased the disk a week ago, immediately I began to receive adverts for it, and they tailed off after a few days. Every single penny spent on that advertising was wasted, I already _bought_ the hard disk. The online grocery store I sometimes use actually builds custom adverts with my exact list of groceries in it, then advertises them to me the day after I order them for delivery, again a human looking at this can see instantly that it's insane, but so long as you don't understand time's arrow (and an advertising algorithm has no a priori reason to do that) then the correlation looks really good...
You make an interesting point about something I thought I understood, but actually didn't.
I always assumed the point of this advertising approach was that it's messy, but it works. Someone who just bought shoes probably won't be buying a new pair ASAP, but there might be better odds than for any random person. Maybe they like shoes, maybe they'll return them, maybe that's their go-to gift for friends.
I genuinely hadn't thought it might just be an exercise in fudging correlation numbers to make ads look more successful than they are.
I bought some very expensive high heels as a treat for myself. They don't make them in my size, I bought them because they're beautiful and put them on display.
In the following weeks I received lots of advertising for those exact shoes. Not for high heels generally, not for nice frocks, and certainly not for glass display shelves, just endless adverts for the same exact pair of shoes.
Somewhere a "digital advertising executive" is telling their client that these adverts were very highly correlated with sales of the shoes. Of course, they'll disclaim, correlation isn't causation. But it is - that's their dirty secret, just the reverse of the causation pretended to. Advertise to people who are already buying the product, then say "See advertising works".
The shoes are the most blatant example because they're so noticeable (I did not spend that much on shoes which are merely a pretty red colour or covered in sparkles) but I even saw this for mundane things like a Seagate hard disk. I purchased the disk a week ago, immediately I began to receive adverts for it, and they tailed off after a few days. Every single penny spent on that advertising was wasted, I already _bought_ the hard disk. The online grocery store I sometimes use actually builds custom adverts with my exact list of groceries in it, then advertises them to me the day after I order them for delivery, again a human looking at this can see instantly that it's insane, but so long as you don't understand time's arrow (and an advertising algorithm has no a priori reason to do that) then the correlation looks really good...