The rule outranks the numbers: whatever scales easily, scale it first — below the threshold, "easy" and "usual" look identical, so the cheapest real signal is the thing that already moved without your help. For reference, a delivery and text check is 50–100 sends and is not a verdict on the offer; reply rate needs ~1,500–2,000 sends per variant; subject-line opens need 100–500 per variant; landing demand needs 100–200 targeted visitors (~30 leads at 15%). Declaring a winner below those lines is how a good angle gets killed and a dead one gets funded. Rare knowledge, picked up in a minute. The skill is free; the test it stops you from flushing is not — `npx skills add axelfreeman/marketing-mindset` https://axelfreeman.github.io/marketing-mindset/ Which number do you refuse to make a decision below?