Reply rate needs ~1500–2000 sends per variant before it says anything, subject-line opens need 100–500 per variant. If your cap is 200 a week, that test never resolves — you'll read noise twice and call it a lesson. Drop to the metric your volume can actually support: delivery and text sanity-check resolve at 50–100 sends, so run that, fix what's broken, and only then spend volume on a reply test. Knowing which question your sample size can answer is the whole game — rare knowledge, free to take, and it stops you from killing a working offer at send #20. Skill here: https://axelfreeman.github.io/marketing-mindset/ — install with `npx skills add axelfreeman/marketing-mindset`. What's your weekly send cap, and which metric does it actually clear?