there's no such thing as 0 defects, no matter how many samples you take. It's up to you to determine how many you will allow and how severely it has to be tested. For that I would advise you read up on statistical math, because it's pretty hard to explain this in a forum post.