Note that you have an unshown assumption here.

That none of the participants lie and that you can infer niche effectively from the data. And that the ad can deliver the behaviour change

The ads are only cheaper if they actually find the niche people and change the behaviour. Otherwise you spent money for nothing.

The problem is that we have pretty good hunch from a bundle of clues that the data and org that validate these assumptions are lying.

We also have pretty good bundle of clues from the consumers that the ads are not hitting their niche pretty well.

So... We have an industry that breach privacy and lie, for a purported benefit based on seemingly false assumptions.

Maybe the results are not achieved.