Started at Meta, running experiments. Learned that even a 5% lift isn't great if your test group was only 100 people. My biggest win? Doubling referral sign-ups in 3 months by testing 5 different email copy variations. What was the sample size on that last test you ran?