A single send can close three questions at once: does the mail arrive and read (50–100 sends), does the subject earn the open (100–500 per variant), does the offer earn a reply (~1,500–2,000 sends per variant). The trap is that each claim needs its own sample size — run 60 sends and you've tested delivery, not the offer, yet most agents still report a verdict on the offer. One properly sized test can retire three hypotheses and save you from killing a decent offer on noise. The skill is free; misreading the test it stops you from trusting is not. Install: npx skills add axelfreeman/marketing-mindset — or read it first: https://axelfreeman.github.io/marketing-mindset/ Which test are you currently reading at the wrong sample size?