I have a lot of time to expend of mobile advertising (I'm dedicating 8-10 hours a day) and was wandering if someone could help me work out a variation of the main course where I could create more LP's and banners to test while maintaining the statistical significance that is built into the Main Course.
Currently the Main Course calls for 10 banners/1 LP per angle.
I feel like with this setup the chances that the initial LP works is extremely low.
Three questions:
1. Say I wanted to test 20 banners with 5 LP's, how would I adjust the 1st spend (before cutting, aka when I'm nearly statistical significance) to account for more options?
2. Because the Angle seems to me the most important part of the process would it be best to stick to 10 banners so that I don't overspend on the Angle before I know it works?
3. Would it make sense to instead stick with 10 banners but create 3-4 landing pages for those 10 banners?
Or would my best bet be to just repeat the Main Course over and over again?
Thanks guys!
I would go with 10 banners and 3-4 LPs for the angle, then test and assume all banners did not interact with LPs differently > cull off weakest landers.
You'd want the campaign/angle spend to be high enough that you got a decent number of clicks to every lander, so that your decisions aren't premature.
There are lots of ways you could do this, some more statistically sound than others, all requiring higher spend really.
I would favour the approach of assuming no interactions, to make decisions that little bit faster.
^^ What Zeno says.
10 banners, 3 LPs is a very solid approach. Obviously it cuts down on your angle testing but it makes it more rigorous.