I just happened across this on the Web, and found it quite interesting.
It's a PDF comparison of a lot of the biggest-name brands' advertising, plus some others that you'll be familiar with (tactical flashlight, anyone?). They look at the ads that ran longest and shortest, as measured by AdBeat - I found this particularly interesting because rather than just cherry-picking a bunch of high-performing ads, they collected the ads run by each brand, and then contrasted the apparently-successful ads for, say, TurboTax, with the less-successful ones.
Brand advertising obviously has some serious differences to affiliate stuff, but still, an interesting Friday afternoon read.
http://blog.capitalandgrowth.org/wp-...nes-adbeat.pdf
This share makes me a bit wet, nice!
Interesting stuff 
They call it headlines test yet the ads are totally different, so it's much more than just the headline. Very good info still but not as "clean" of a comparison as I would like to see 
Edit: LOL, point in case right here...

Now this is a very nice report, always great to see the results of real campaigns, thanks for the share Hugh 
Thanks Caurmen, very "easy on the eye" report.
Their methodology is flawed, and their analysis is based on assumptions.
"* We used AdBeat to assign a score to the banner ads based on their frequency and longevity. Advertisers usually test different ads and then let the top performing ones run the longest. No one runs ineffective ads for a long time."
You would think that is the case...but it's a MEGA assumption.
You could also get max longevity points by spending $5 a day.
Glad you're finding it interesting, everyone!