Since you used a beer example however, I will point out that Claude Hopkins famously made Schlitz beer #1 by using modern, trackable, results-driven advertising that he split tested relentlessly. This was close to 100 years ago, so I guess not much has changed(or brand advertisers haven't evolved much in the last century).
So the A/B test you want can't be something as simple as: we show user A a 7-Up ad and don't show it to user B, and see if that changes 7-Up purchasing habits.
Rather, it's something like: in the year 2011, spend $m on brand ads trying to improve the image of 7-Up, so people think it's cool and not the red-headed step-child version of Sprite. Then, re-run that same year 2011 without spending the money, and see how much difference there was in total sales.
The problem, of course, is that it's really hard to rerun years. ;-) You can try it in different years, e.g. increase or decrease your ad budget in 2012 relative to 2011, and see if it has an effect. But then you have to find some way to control for the fact that 2012 is different than 2011 in a lot of other ways too, not to mention that it follows 2011 so isn't independent of it, which makes it hard to draw any firm conclusions about which level of ad spending had what effects.
Of course, you could argue that those vague branding goals aren't particularly important, but I think in a lot of areas they are: the reason someone like Corona or Budweiser sells as much beer as they do is largely because of those brand ads that don't necessarily directly lead to sales, but build an image around the brand that lets them sell more beer for more money than they'd be able to do if they were unknown Brand X.