Bayesian A/B Test Calculator

Stop guessing if your A/B test results are significant. Get a complete Bayesian analysis with probability to win, expected lift, risk quantification, and actionable recommendations - all calculated instantly in your browser.

Probability analysis
Chance each variant beats the other
Lift estimation
Point estimate & 95% credible interval
Risk quantification
Expected loss for each decision
Clear recommendations
Implement, keep, or continue testing
SRM detection
Sample ratio mismatch checks

How it works

Bayesian Approach

Uses Beta-Binomial conjugate priors for exact posterior distributions. No p-hacking or multiple comparison issues.

Monte Carlo Simulation

100,000 samples from posterior distributions ensure accurate probability estimates.

Expected Loss

Quantifies the cost of making the wrong decision, helping you make risk-aware choices.

Built for marketers, product managers, and data analysts who want statistically sound A/B test analysis without the complexity.

Enter your test data

Input visitors and conversions for each variant

Control Group

Variation Group

Frequently Asked Questions

Why Bayesian instead of frequentist statistics?

Bayesian analysis gives you direct probability statements like "There's a 95% chance variation beats control." It also allows for continuous monitoring without inflating false positive rates, and naturally incorporates prior knowledge when available.

What does "expected loss" mean?

Expected loss quantifies the potential cost of making the wrong decision. If the expected loss of implementing the variation is 0.05%, it means on average you'd lose 0.05 percentage points of conversion rate if the variation is actually worse. This helps you make risk-aware decisions.

What is Sample Ratio Mismatch (SRM)?

SRM occurs when the actual traffic split differs significantly from the expected split (e.g., 50/50). This can indicate bugs in your experiment setup, bot traffic, or data collection issues. The calculator checks for SRM automatically and warns you if detected.

How much data do I need?

There's no fixed minimum, but Bayesian analysis becomes more reliable with more data. As a rule of thumb, aim for at least 100 conversions per variant for meaningful results. The calculator will suggest continuing testing if there isn't enough data to make a confident decision.

Is my data stored anywhere?

No. All calculations happen entirely in your browser. Your data never leaves your device and is not stored or transmitted anywhere.