Another common bias is if your are running multiple split tests in parallel you need to assign each visitor to an option in each split test at RANDOM. If you do

Another common bias is if your are running multiple split tests in parallel you need to assign each visitor to an option in each split test at RANDOM. If you don't you'll end up with 50% of the users going though all the same 50% of the options and the other 50% of users going through the other 50% of the options.