Primary — what actually gets done
The test plan: four candidates, chosen
Run simultaneously. $25/day each for six days. Each is a different kind of bet, so whatever happens you learn something that generalises.
Before launch — build once, reuse four times
Pre-committed thresholds — decided now, not after
The goal: a business with repeat customers
Assume several will fail — the plan budgets for it
Testing budget
Supporting — candidate supply
Research pipeline
Cheap to run and worth keeping deep, since failed tests need replacements ready. Not the finish line.
Supporting — runs automatically
Research runs
Every Monday at 8am, automatically
Highest scoring
What the rubric ranks first — deliberately not the same as what gets tested first
Deliverables
Everything produced so far
Open issues and decisions
Things that need you, not the loop