On top of that, I'd wager most people don't very strongly weight a Youtube advertiser with the video it's played before. At least not me.
Meanwhile, like you said, I’m served the same handful of annoying ads over and over, for products I’d never consider buying. Where are all the ads for stuff that’s so desirable and targeted to me that I just can’t resist purchasing them??
Not saying that critics of advertising have no valid claims, but the juxtaposition is funny.
That said, I think the reason why machine-driven advertising is so much better than its predecessors is simply that the predecessors were really awful. They used to do market research by installing double-sided mirrors in department stores then taking notes on what people looked at. It was absurd. Now, there's actually an even-odds chance that a targeted advert will meet somebody that might be vaguely interested - and that's absolutely orders of magnitude better than a billboard.
And also, to be honest, this whole complaint about cost-performance indices is a bit absurd - since as far as I can gather, they were basically oneiromancy before Google and Facebook came along.
I'm not so sure we have progressed that much since then. You search for something on Google, swap to Facebook and an ad appears for that same exact thing on the sidebar - not really any different than using browsing as a determining factor for desire to purchase.
A small aspect of ads, I know but I found it such a ridiculous truth that I can't imagine the rest of their systems are actually all that smart.
To give you a perspective, efficient "performance" (the kind which wants you to buy now) advertising is a click ratio of about half a percent on ads, which vastly outperforms the basic performance one, usually having a ratio of about a tenth of a percent. Of those fractions of a percent, a fraction of a percent will actually buy...
That still means ads missed the target 995 times out of 1000, so are still globally horrible.
Nevertheless, since ads on the internet are so cheap (prices are often negotiated by 1000 ads, to make it a bit more tangible, and we're talking like 5$ per 1000 ads here), it's still sometimes worth it.
So yeah, the global experience is crap, but that's often not the point anyway.
https://www.goodreads.com/book/show/3730.The_Hidden_Persuade...
Or worse, for big-ticket items that you'd only buy infrequently. HEY THIS GUY BOUGHT A DRONE LETS SEND HIM DRONE ADS! WAIT NO NOW HE BOUGHT A CAR HE'S A CAR BUYER SEND CAR ADS. No, you silly equation, I just bought a car, that means I'm NOT in the market for a car.
Then they sell us household gadgets to sink their hooks deeper into our lives. Microphones listening to conversations won't be enough; soon we'll herald the coming of new "smart" appliances like the Google Toilet, which plays your favorite music, recommends better posture, sprays you off when done and stealthily analyzes your stool samples for markers to see if the Captain Crunch ads they bombarded you with over the last week led to a conversion.
It doesn't work, but failure is clearly no impediment to progress.
The reason this happens is that nobody involved gets anything out of stopping this happening. Incoming extreme cynicism.
First off, the marketers are still mostly using demographic segments - often a marketer will encounter a machine-learning model, and their first question is "who are these people" and when you can't explain how the model works, they revert to using demographic segments because they trust them more. This means they often do analysis demographically, and over target people in particular narrow age and gender demographics because they did better last month by chance.
Second, management want to keep their budgets large so their resumes look good. Their desire is to increase the amount of people they can target, while simultaneously appearing to work on interesting tech. Since social and video have recently been en vogue, spend increases on these platforms, while projects focus less on increasing spend efficiency and more on taking credit for sales by tracking customers through the pipeline. This creates a weird scenario where everyone says they're using sophisticated machine learning models, but the contractors and consultants involved are glorified powerpoint factories.
Third, the big silos, google and Facebook, probably have some incredibly sophisticated models of consumer behaviour that are incredibly efficient - but why would they want to give their advertising customers something that would reduce spend?
I'm sure it raises "brand awareness" measurements, at the cost of making people hate them.
The reality is that people don’t actually want enough to cover ad revenue, but ads make them buy more.
The rule is roughly "narrow targeting, cheap impressions, reasonable frequency, pick any one". A shady ad agency (but I repeat myself) will never pick "reasonable frequency" because their advertisers aren't sophisticated enough to ask if their ad is being sent to one person 50 times in a day.
It only takes one buyer to not set the frequency control option to spam you with the same ad over and over -- everyone else is one-and-done.
Ugh. Seriously though I suspect it is because their algorithms have decided that you like X and therefore they only show you adverts about X, which is a small subset of all available adverts, so you get the same ones again and again.
On TV they can't target so accurately so you basically get shown every advert. This means there are more to show and it is less repetitive.
I also think advertisers don't mind showing the same person their advert a million times. It may be annoying but it works - name one competitor to Grammarly off the top of your head.
I'd go as far as to say that the end user would be more receptive to ads, if they filled ads with pro-bono charity ads promoting some charity or another ... then the user wouldn't start to hate ads, and they could probably write off the ads they donated.
Similarly, twitter showing ads for Mac/windows only product when my user-agent clearly says Linux!