Stellar answers to this line of questioning will consider the use-case scenarios and demonstrate a basic understanding of manufacturing variability and statistics. This is a very revealing question, the why being more important than the number. Except in extreme situations, any answer that is less than five or more than 100 is probably wrong.
A more in-depth answer can be found here! Mainly depends on the desired percentage of confidence, desired percentage of reliability, and the type of data being collected from the samples.