You're using the NORM.DIST function incorrectly. The first argument is the value whose cumulative probability you want to determine, which should be either 9.8 (to get the probability that a sample will be undersized) or 10.2 (to get the probability that a sample will not be over-sized). From that you should be able to determine the probability that any one sample is defective. Then the next step is to determine the mean and standard deviation for the expected number of defective parts in a batch of 1000 (you should be familiar with the binomial distribution). And then finally you should be able to determine the probability that 50 or more parts would be defective out of the batch of 1000. Hope this helps.