AB Test Sample Size Calculator

AB Test Sample Size Calculator MCP Connector for Claude

A+

Calculate required sample size, test duration, and peeking risk for A/B experiments.

3 tools Official Updated Jun 29, 2026 Official Vinkius Partner

This MCP server provides essential statistical tools for running reliable A/B tests. Use calculate_required_sample_size to determine the number of users needed per variant based on your baseline conversion rate and desired Minimum Detectable Effect (MDE). With estimate_test_duration, you can project how many days your experiment must run given your daily traffic. Finally, use assess_peeking_risk to evaluate the danger of stopping a test prematurely, helping you avoid false positives caused by early analysis.

ab-testingstatisticsdata-scienceexperimentationconversion-rate

3 tools expose this connector's capabilities to your AI agent.

calculate_required_sample_size

Determine the required sample size for an A/B test

estimate_test_duration

Estimate the duration of an A/B test

assess_peeking_risk

Assess the risk of stopping an A/B test early

See how to talk to your AI agent using AB Test Sample Size Calculator.

I have a baseline conversion rate of 10% (0.10). I want to detect a 5% relative change (0.05) with 95% confidence and 80% power. How many users do I need?

To detect a 5% relative change from a 10% baseline with 95% confidence and 80% power, you will need approximately 156,432 users per variant, for a total of 312,864 users.

If I need 50,000 total users and my site gets 2,500 visitors per day, how long will the test take?

Based on a requirement of 50,000 users and daily traffic of 2,500, your test is estimated to run for 20 days.

My A/B test has been running for 5 days. The planned duration was 14 days. What is my risk level?

With only 5 days elapsed out of a planned 14, your peeking risk is High. You should continue the test until the planned duration is reached to avoid false positives.

You can use the `calculate_required_sample_size` tool. By providing your baseline conversion rate, MDE, confidence level, and statistical power, it will return exactly how many users you need per variant.

Related Connectors