[antonfire]

> Advertisers will produce statistics to prove that their kind of marketing "works". But how could one prove such a thing? It would involve peering into people's minds.

Or, say, running an A-B test and measuring whether people in the control group buy more of the product or feel more positively about the brand than the exposed group.

[denton-scratch]

An A/B test won't tell you anything about how people feel about a brand. A/B testing is useful for telling you how badly your UI designers screwed up. It's a way of testing the effectiveness of websites.

It won't tell you anything about the usefulness of plastering the world with billboards, nor of TV advertising.

[antonfire]

A-B testing will tell you whether internet advertising works.

Pick an ad. Assign users at random into groups A and B. Show users in group A the ad. Don't show users in group B the ad. Watch and see at what rate users in group A and in group B buy the product. (Or survey the users in the group to ask how they feel about the brand.) If there's a statistically significant difference between the behaviors of people in group A and in group B, then you have statistically significant evidence that the ad works.

Yes, this kind of thing is harder to do with billboards and with TV advertising than with internet ads. This is one of the selling points of internet ads over TV and billboards. If you buy ads from Google or Facebook or whatever, they can run this kind of experiment to measure how effective your advertising actually is. It doesn't involve peering into people's minds, just watching their internet behavior and/or surveying them.

Relevant links:

Google help page: https://support.google.com/displayvideo/answer/9570506

Facebook help page: https://www.facebook.com/business/help/1693381447650068