I work at a taxi company. Recently I have discovered errors in the assignments. Errors in assignments seem to show up randomly. There are more than 800k driving assignments yearly. There is no way I could sample all of them. Since the assignments can be considered random I was thinking of sampling based on what I experience in my assignments, since the assignments are random. The exact same error is experienced randomly by other drivers as well. Would it possible to measure an overall value ( mean value ) based on my own assignments with 95% confidence and level and +-3% error, given large enough sample size as measured in my assignments?