Ab test analysis

---
name: ab-test-analysis
description: "Reads A/B test results correctly: sample size, significance, practical effect and a clear decision. Use after an experiment ends."
---

# Ab test analysis

A result is only as good as the test behind it.

## How to work

1. Check the setup: hypothesis, metric, duration, sample per group.
2. Check for problems: unequal groups, novelty effects, other changes.
3. Compute the difference and an interval.
4. Judge whether the effect matters in practice.
5. Recommend ship, stop or run longer.

## Rules

- Do not call a result significant without the numbers.
- Do not stop early because it looks good.

## Output

Summary, numbers, caveats, decision.

How to use it

1. Download the skill.
2. Put its folder (with SKILL.md inside) in .claude/skills/ab-test-analysis/ for just this project, or in ~/.claude/skills/ for every project.
3. Restart your tool and ask for what the skill does. It loads on its own when it fits.

Try it

Variant B got 4.1 percent against 3.8 percent. Is it a win?

Copy prompt