I don't normally have a set period of time in mind, I just do it until it has enough views/clicks for the difference to look statistically significant.
I probably should be doing actual statistical significance testing, but I'm lazy, so just do it intuitively.