I'm setting up a new campaign for adult after getting my feet wet with a little test over the last week or 2. My clusterfuck of variables will look something like this:
2 or 3 offers
2 or 3 landers
3 angles
3 banner templates
5 placements
8-10 images per placement
Now, I imagine this is what a typical campaign looks like for most of you guys, but how do you test everything? Is it best to keep the backend of the campaign (offers and landers) the same across the board, and change only the banners? Maybe it's better to keep each campaign for each placement exactly the same or 100% different?
If we take the landers as an example, I could either test 3 landers across 5 placements, or I could test 15 different landers (3 per placement). I'm testing on a fairly tight budget (roughly $3k) so I want to make sure I achieve statistical significance.
Testing completely different angles and landers across the board feels like throwing shit at the wall until it sticks, but testing only 3 angles and 5 landers is limited and is a waste of money if none of them convert well.
Maybe a mixed approach like 2 landers kept the same across the board and 1 different lander per campaign would be better? (eg. In campaign 1 you would have lander 1, 2 and 3, in campaign 2 you would have lander 1, 2 and 4, in campaign 3 you would have lander 1, 2 and 5 etc.) Please let me know what you guys think.
Assuming that you are new to the GEO / traffic source, I'd say that 5 placements are too much to start with.
Offers and LPs usually perform pretty consistent for me over different placements, especially on the same traffic source. I'd focus on 1-2 high volume placements, find the top offer/LP combo and then expand to other sources.
I know, ideally you'd want to test every possible combination, but especially with a smaller budget I think it's better to limit the variables a bit.
Obviously I'm not the most experienced person hear, but I remember reading that the two mistakes people make when running campaigns are having one ad, and having more than two ads.
Maybe this only applies to Ad Groups and Search, but my initial reaction to split testing is 'test EVERYTHING because I can'.
Instead of doing that, I'd say that you would probably be better off testing two to three ads in a single position with two offers. This would give you 4-6 total variables which should allow you to reach significance really quickly. Once you've culled a few options, like say that you found the best ad, then you could split test placements or images a bit more.