I know no one is going to 100% reveal what they do for split testing, but I'm stuck not knowing whether to spend lots of time and money testing each variable thoroughly vs throwing shit at the wall until it sticks for lack of a better name.
What sort of approach produces the fastest/best results?
Approach A: testing lots of varied banners with big differences and hoping one sticks. It looks like this is something maynzie has done here if he tested all of those images at the same time: http://stmforum.com/forum/showthread...-(With-a-Twist)
Approach B: More methodological, but testing multiple things at once. I see this one suggested a lot. If you test 20 images, 2 templates and 2 angles each, you have only 5 images per template/angle combination. Enough to offset the affect which the image variance has on the stats?
Approach C: Something I've experimented with. Taking 2 good converting images and testing them with 3 angles and 3 templates. Limited and time consuming.
What are your thoughts with this?
Firstly, I'd always run one angle per campaign. Your LPs need to be hooked into your angle anyway, as much as your banners.
I normally then test images, then headlines. Images are by far the most important thing to a banner in my experience - that might vary depending on vertical, but I can't off the top of my head think of one where images aren't so important ( financial, maybe? ).
I normally test templates once I've got some working images and need to start combating burnout.
Hope that helps! I'm sure we all have different approaches...