In my Intro survey I have an 18% branch – At first it seemed to be working at about the right pace. For the last few months it has been pulling more like less than 5% of the people? Any ideas?
SG’s documentation about the percent branch talks about the way it works (http://help.surveygizmo.com/help/article/link/branching-a-b-split-testing), I’m not sure if you’ve run across that. It looks like it’s not abnormal for the percentages to be off since the whole thing is completely random, like a coin flip. The way to increase the odds of the branching percentage being more of what you expect would be to receive more responses, if that’s a possibility for you.
There are tons of ways how to maintain even split in A/B test rather just relay on PHP rnd that is not supposed for use as “end-user” function. Current branch is mostly useless (already many years!!!) exactly because you never know if sample inequality will be acceptable or terrible.
This is extremely pity that SG ignores repeated users request….