Banners - Display campaigns and Remarketing

Creating separate ad groups for various banner sizes

[SEMblahblah]

Is it a good idea to create 2-3 ad groups with the same audience/keywords containing different banner sizes that you'd like to promote?

The point being - all sizes perform differently and I'm looking at the best way to get maximum sales out of each size while maintaining the CPA. With all banners in one ad group, if I reduce the bid to control the CPA, it would have an impact on sites that my top banners are placed on which is not desired. I'm planning to club 2-3 banners together in one ad group, say an ad group called top performers, which would have a higher bid than banners in other ad groups and so on.

I'm currently planning to try this with a remarketing campaign since there'd be no keywords in there. But would it work differently with a regular keyword+placement display campaign where each of these ad groups would have the same set of keywords? I guess they'd be competing against each other and then the ad with a better ad rank would show.

I'm sure people must've tried this. So, can anyone shed some light on the pros and cons of this strategy?

Thanks!

[RhinoFish]

If it's super niche, banner fatigue will make your multi strategy work at times. But if it's widespread targeting, you'll see banner rotation in Display where they're limiting the repeat views, though a lesser performing ad gets into the rotation. So for the effective output to happen, I think the key question is, do you have the resources free to make the extra work for managing this, and is the marginal increase in return worth that incremental opportunity cost. Generally, it isn't. Then again, my general and yours are different - I generally see other opps within an account that have a higher ROAS priority than segmenting banners in ad groups for rotation purposes. Sorry, that was a long non-answer. :-)

[SEMblahblah]

Totally get what you're saying. I'm doing this only for one of my campaigns which contributes to around 50% of overall conversions. So, if I'm able to optimize it 'just a little more', it'd be worth the time in this case.