In practice, doing this will lead to what the author talks about under "Stopping tests as soon as you see winning results will create false positives." It doesn't make your data invalid, but it will generally lead to poor methodology if you are not careful.