Starter prompts
4 ways to start with Experiment.
Hypothesis
→ Tight
▸ Preview prompt
Tighten this experiment hypothesis: name a single primary metric, an expected effect size with rationale, a stop rule (sample size + minimum runtime), and one guardrail metric. Then list the 3 most likely confounders.
Loose hypothesis: 'Adding a tooltip on the pricing page that says "Most teams pick Pro" will increase Pro-plan signups.'
Context: ~12k pricing page visitors/week. Current Pro signup rate is 3.4% of visitors. We've never run an A/B test on pricing copy before.
Power
→ Sample size
▸ Preview prompt
Calculate sample size + duration for a checkout A/B test on 8% baseline conversion and a 5% relative lift target.
Decision
→ After result
▸ Preview prompt
Result came back inconclusive. What's the decision process — extend, kill, ship anyway? Walk me through it.
Library
→ Build memory
▸ Preview prompt
Set up an experiment learning library so we stop running the same test twice and remember what didn't work.
What it does
Tasks Experiment ships every week.
Design
- Hypothesis + metric
- Sample size + power
- Variant + randomization
- Pre-registered analysis
Tracking
- Result dashboards
- Decision log
- Learning library
- Stop rules + holdouts
Worked sample
A real Experiment chat.
Pairs well with