Bayesian updatingBeta-binomialA/B testingMotion for React

Bayesian updating for an A/B test.

Watch a shared Beta(3, 17) prior absorb conversion evidence from two variants and become two posterior beliefs about the true conversion rate.

View work Read writing

Posterior lift

Variant B moves the posterior mean to 17.2%, about +4.7pp above Variant A after observing the experiment.

Bayesian A/B update

Prior belief plus conversion evidence becomes a posterior.

The same starting belief is applied to both arms. Running the update adds conversions and non-conversions to each Beta distribution, then the curves separate into Variant A and Variant B posteriors.

Conversion rate belief

Beta(3, 17) to Beta posteriors

Starting point

Both arms share one prior

Run the update to add observed conversions and non-conversions to the prior.

Prior

15.0%

Before evidence arrives, both variants share the same Beta prior.

Evidence stream

Visitors land, conversions light up, and each arm updates its Beta posterior.

Probability B beats A

Run the update to compare posterior draws.