I want to answer {{question}}. What's the right way to test this?
Context:
- What I have access to: {{data, ability to run experiments, surveys, etc.}}
- Time and budget: {{realistic constraints}}
- Stakes: {{how confident do I need to be?}}
Output:
1. **The hypothesis** in falsifiable form — phrased as something that could be wrong
2. **What evidence would confirm** vs **what would falsify** — be specific about the threshold
3. **The right test** — A/B test? observational study? survey? lookback analysis? back-of-envelope?
4. **Sample size or data needed** — rough estimate of what's "enough" to be confident
5. **The faster, cheaper version** I could run first as a smell test before investing in the rigorous version
6. **What confounds I need to control for** — and how
Be honest if the question is hard to answer rigorously. "You can't, but you can get a directional read with X" is a useful answer.hypothesistestingexperiments