Facebook now offers the ability to run rigorous controlled experiments on your advertising spending - you basically randomly divide your audience into treatment and control, run an advertising campaign that only the treatment group can see, and compare purchasing behavior between the groups.

On the basis of several experiments with this, I can assure you that advertising does not always work, even if you track the treatment group for months after they were exposed to ads.

I think the author's bigger point is that a lot of the purpose of advertising basically boils down to increasing brand recognition. That's fair, but not anywhere near as novel of an observation as he thinks it is. His assertion that the majority of advertisers don't understand awareness and brand recognition is laughable - they understand it very well, and consistently run campaigns to achieve it.