Tools

Bayesian AB Test Calculator

Calculate the probability of one variant beating another using Bayesian inference with Monte Carlo simulation. Choose your model below depending on your metric type.

Bernoulli-Beta is for binary outcomes: did the user convert or not? Use it for conversion rates, click-through rates, sign-up rates, or any yes/no metric. Each user is one trial with two possible outcomes.

Control (A)

Users / Sessions

Conversions

Variation (B)

Users / Sessions

Conversions

Prior Parameters

Alpha prior

Beta prior

Common prior configurations

Scenario	Alpha	Beta	Why
No prior knowledge	1	1	Flat prior. The posterior is driven entirely by your test data.
Landing page ~3% CR	1	39	Light prior based on ~40 virtual users. Washed out after a few dozen real visitors.
Email opt-in ~6% goal	12	188	Moderate prior (~200 virtual users). Smooths randomness in early opens and clicks.
+25% uplift over 4% baseline	4	76	Weak optimistic prior anchored at 5%. Yields quickly as real data streams in.
Strong historical data (8% on 10K visitors)	801	9201	Empirical Bayes. Like adding one extra day of traffic to a trustworthy baseline.