Quality Digest

I generally use a sample size of n=30 to perform a capability study on a new product. I am asked frequently, why is 30 the magic number. I do not have a statistical explanation of using n=30 or how much confidence is lost by using a smaller sample. I usually state that 30 is a good number for estimating normality. If the capability results are from destructive testing or very costly testing, what are my risks of going with a smaller sample to measure capability and how can I quantify those risks?