Ab Test Designer

A/B Test Designer

Overview

Designs rigorous A/B tests with clear hypotheses, variant specifications, sample size calculations, and analysis plans.

Workflow

Step 1: Define the Test

1. What are you testing? (page, email, ad, feature)
2. Current metric: What's the baseline performance?
3. Goal: What improvement are you hoping for?
4. Traffic/volume: How many users/emails/impressions per day?

Step 2: Structure the Test

Output Format

# A/B Test: [Test Name]

## Hypothesis
If we [change X], then [metric Y] will [increase/decrease] by [Z%] because [reasoning].

## Test Details
- **Type**: A/B / A/B/C / Multivariate
- **Primary metric**: [what you're measuring]
- **Secondary metrics**: [supporting metrics]
- **Guardrail metrics**: [what shouldn't get worse]

## Variants

### Control (A)
[Description of current experience]

### Variant (B)
[Description of the change]
[Mockup/wireframe description if applicable]

## Sample Size & Duration
- **Baseline conversion**: [X%]
- **Minimum detectable effect**: [X%]
- **Statistical significance**: 95%
- **Required sample size**: [N per variant]
- **Estimated duration**: [X days]

## Analysis Plan
1. Wait for minimum sample size before checking
2. Check primary metric first
3. Segment analysis: [segments to check]
4. If significant: [implementation plan]
5. If not significant: [next steps]

## Risks & Considerations
- [Risk 1]: [mitigation]
- [Risk 2]: [mitigation]

Quality Checklist

• [ ] Hypothesis is specific and falsifiable
• [ ] One primary metric defined
• [ ] Sample size calculated
• [ ] Duration estimated
• [ ] Analysis plan prevents peeking bias
• [ ] Guardrail metrics identified

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: lionelsimai
• Source: lionelsimai/claude-skills-collection
• License: MIT

Install and usage instructions live in the source repository linked above.