Lately I've started to realize the importance of testing EVERYTHING- offers, landers, offer LPs, creatives, etc.
I use imobitrax which makes split testing a breeze.
Right now, I'm trying to develop a reliable split testing process to optimize.
Here's what I'm thinking (using landing pages as an example):
1. Test 4 completely different styles (ex. flog vs. quiz vs. chat etc.)
2. Take the winner and create 2 variations with different copy/images and test those
3. Run 50% of traffic to the winner. Add 1-2 new styles to the testing rotation per week and repeat
Those of you who test a lot, what are your thoughts on this?
Also, how do you deal with anomalies in your data?
Like I'll have 5 practically identical campaigns running on the same traffic source. In 4/5 of the campaigns, one particular offer/landing page will dominate the other ones. But in one campaign, the complete opposite is true and I have no idea why. (Btw, I always make sure my data is statistically significant with splittester.com.) I'm running on mobile, so maybe I just need to look at data from longer time intervals.
I would push 80% traffic to the winner... otherwise you'll be losing more money than you might like. 20% is enough to get a good idea of performance, and if it's in the ballpark, hit the 50/50 switch for any LP that looks like a serious contender.
The anomaly thing is just a mindfuck, but it doesn't matter that much, especially if you've used FB in the past... If you split a dataset into five separate ranges, not each of those ranges will allow you to make the same conclusions. You already know the big picture so stick with it... Maybe you're being assigned different quality/segments of traffic based on platform-specific factors when creating those campaigns..
old school but the concept still works :
http://www.splittester.com
we have lost a lot of cash thinking a Loser is a Winner...
main point:
find out the TRUE certainty of your ad,
ALLIANCE