A/B testing
Only test one dimension at a time (subject line, opener, or CTA) and wait for meaningful sample size.
Decision rule
Avoid false winners
Do not end tests too early. Keep both variants running until you have enough replies to reduce random swings.