Actually designing those tests is easy. The trouble is that in order to run them you need to be tracking data over a long period of time, and executives lack the necessary attention span.
The second problem is that one of the costs of excessive advertising is to your brand. And testing things that affect your brand is something that A/B testing is not good for.
There probably have been a lot of models showing that asset prices will not go down more than x%, because they have not gone down more than x% anytime in the last n years for which we happen to have data.
Similarly, 17 ads on a page might maximize eCPM today. But it is also possible that showing 17 ads on a page today could lead to fewer users visiting the site tomorrow. In other words, it is difficult to design an A/B test that takes into account long term effects on your business.