We have written a bit on sample size for common events. We would like to extend this analysis to rare events. In web marketing and a lot of other applications you are trying to estimate a probability of an event (like conversion) where the probability is fairly low (say 5% to 0.5%). In this case Related posts: