Design A/B tests with proper methodology, sample sizes, and success criteria.
/experiment-designer2-3 hrs → 15 min
Compared to doing it manually
/experiment-designerType this in Claude to run the skill
Shipping changes without testing is risky. Poorly designed tests produce invalid results.
Agent workflows chain multiple skills into one command.
.claude/skills/ folder in your project/experiment-designer in Claude to run the skill/metric-framework-builderBuild comprehensive metrics frameworks using the AARRR pirate metrics or input/output methodology.
/funnel-analyzerDiagnose conversion funnel problems and generate data-backed improvement hypotheses.
/ab-test-analyzerInterpret experiment results with statistical rigor and clear ship/no-ship recommendations.
/ab-test-designerDesign statistically sound experiments with clear hypotheses and sample size calculations.
A/B tests compare two versions of the same thing. Experiments are broader — they can test hypotheses, validate assumptions, or explore new directions. All A/B tests are experiments, but not all experiments are A/B tests.
Start with a hypothesis ("We believe X will cause Y"). Define success metrics before you start. Minimize variables to isolate cause and effect. Set a timeline and commit to acting on results.
Failed experiments are successful learning. Document what you learned, update your assumptions, and decide: iterate, pivot, or move on. The only failed experiment is one you don't learn from.
Run this skill inside your PM Operating System, or download it on its own.
Use all 70 skills, workflows, and sub-agents in a system that knows your company, product, and customers.