Website, email and ad tests. Know how long to run it, whether it won, and when to stop.
Confidence: how sure you want to be that a winner is real and not luck. At 95%, about 1 in 20 "winners" in tests with no real difference would be false alarms.
Power: the chance your test spots a lift if it really exists at the size you entered. At 80%, a real lift of that size is missed about 1 time in 5.