Silly question, but how much more data do you have in production compared to your test environment? Is your test data obviously fake test data such as lots of similar names, etc?