* A flat probability distribution for "will take between X and Y" is probably wrong, a bell curve would be more accurate (where X and Y are the 95% percentile).

Another question: is it open-source? I'm sure a few people around here would be glad to hack into the code ;-)