A/B Testing
Plans, designs, and implements A/B tests and growth experimentation programs.
Measurement & Testingmeasurement-testingstandard
by AgentVoltv1.0.0Published 1y ago1 install
Free to sign up · every skill included with AgentVolt Pro
Before & after
Same prompt, same agent — the only difference is the skill.
Prompt“Design an experiment to increase signup conversion”
Beforegeneric AI agent
A/B testing signup flow is smart. Before running a test, I'd want to understand your current conversion rate and sample size to make sure the experiment is statistically valid. What specific element are you testing—button color, copy, form fields, or something else? What's your traffic volume per day? Do you have a hypothesis about what would improve conversions? How long are you willing to run the test? Statistical power matters here, and I can help you calculate sample size, but I need more context about your baseline metrics.
Afterwith A/B Testing
A/B TEST PLAN: SIGNUP CONVERSION OPTIMIZATION HYPOTHESIS If we shorten the signup form from 6 fields to 3 (email, password, name only), then conversion rate will improve from 4.2% to 5.8% (+38% lift) because users abandon long forms. TEST SETUP Control → Current signup (6 fields: email, password, name, company, role, phone) Variant → Shortened signup (3 fields: email, password, name) Secondary variant → (Skip: company role ask until post-signup) STATISTICAL REQUIREMENTS Current conversion → 4.2% (baseline) Desired lift → +1.6pp (38% relative lift) Daily traffic to signup → 2400 users Statistical power → 90% (industry standard) Significance level → 0.05 (95% confidence) Sample size needed per variant → 4,200 conversions total Estimated test duration → 21 days (at current traffic) METRIC DEFINITIONS Primary metric → Signup completion rate (submitted form / visitor) Secondary metrics: • Form abandonment rate (started but didn't submit) • Time to complete (seconds from form load to submit) • Email verification rate (clicked confirmation link) • Signup-to-first-login rate (onboarding completion) SUCCESS CRITERIA Win threshold → Variant exceeds control by 1.6pp with p < 0.05 Stall threshold → No significant difference after 21 days Loss threshold → Variant drops below 3.5% (stop test immediately) RUN SCHEDULE Start → Monday (avoid weekend noise) Duration → 3 weeks (21 days) Daily check-in → Monitor for bugs, traffic anomalies Mid-test analysis → Day 10 (readout only, don't act) Final analysis → Day 21 (decision point) IMPLEMENTATION CHECKLIST Code → Variant form deployed to production, 50/50 traffic split Tracking → Events fired: form_started, field_completed, form_submitted QA → Test both paths before launch, verify tracking fires Notification → Slack alert if conversion drops 10% in variant RISK MITIGATION Monitor email delivery → Verification emails sent for both variants Load testing → Variant handles 1200 req/sec (peak load) Rollback plan → Kill variant within 5 min if >20% failure rate detected POST-TEST DECISIONS If variant wins → Deploy to 100%, estimate annual impact ($X revenue) If tie → Run secondary test on field order or copy variation If variant loses → Document learnings, try different hypothesis next month ILLUSTRATIVE NUMBERS: Not investment or business advice.
About this skill
name: ab-testing description: Use when Plans, designs, and implements A/B tests and growth experimentation programs.
A/B Testing
Guides setting up a valid experiment — hypothesis, sample size, success metric, and test duration — rather than just running a coin-flip test.
What you get
- A structured experiment plan ready to implement.
Customize your output
- Which metric to optimize
- traffic split
- test duration.
Example output
A written test plan with hypothesis, variants, and success criteria.
Best for
Marketers and growth teams running structured experiments.
SKILL.md preview
SKILL.md
---
name: ab-testing
description: Use this skill when planning, designing, or implementing an A/B test or a broader growth experimentation program.
version: 1.0.0
category: Measurement & Testing
author: AgentVolt
license: proprietary
tags:
- measurement-testing
- standard
---
# A/B Testing
Sets up a statistically valid experiment from hypothesis to test duration, rather than a coin-flip test with no design behind it.
## When to use
… (sign up to view the full skill)Sign up to view, copy, and install the full skill