They always trumpet this stat but they never address the issue that a ton of banner campaigns are behaviorally targeted, sometimes called "lead back". This means ads are only shown to people who have already been to the website. So going along these lines for these types of programs you could also say that people who previously visited an advertiserís site are more likely to visit that advertisers site again. Then the question becomes does the banner have anything to do with it or would the person have gone to the site anyway? I've never seen any study that addresses this, ever.
I'd agree to a point, but and it's a big BUT, is it reasonable to say that our brains are operating differently when we are browsing the web than viewing ads in ANY/EVERY other medium?
When you read a newspaper or magazine, it tends to be in a more relaxed state of mind, when you watch TV you tend to be vegging out in a relaxed state of mind, when you are driving down the highway and look at a billboard there are not a lot of distractions if the road is relatively clear.
When you go online, you usually go there for a purpose to do something, to achieve something. You probably have multiple windows open, you are more likely to be sitting on the edge of your seat, and your heart rate might even be elevated a bit compared to reading a book or watching a mellow TV program. You are trying to be more focused and yet you have a lot more distractions so intuitively, I would think you pay a lot less attention to advertising unless it is what you are looking for. Since banner campaigns really can't judge or target intent and people are more focused on searching and scanning, I think the branding effect of banners is highly questionable. There have been many brands built on search or search/affiliate but is there even one that was built (even in large part) on banner ads? Videos maybe (think BlendTec), but not banners and probably not rich media either.
There is so much money and there are so many vested interests in claiming banners work but nobody ever tries to prove they don't. There is always some new study that says even though nobody clicks on them, they are a great branding tool and maybe they are but the assumptions and the targeting behind the campaigns need to factor into these studies to make them hold water IMHO.
Let's see a big ad network or ad serving company run a large campaign for a big merchant. Target it however you want but serve the actual banners 50% of the time and just set a cookie as if people have been exposed to the banners the other 50% of the time. Compare those two pools and see what the results show. Chances are the will be very different based on the targeting options and may show that some targeting options work and some don't.