Bayesian

A statistical approach based on the use of probability to model the uncertainty surrounding the parameters of a model. Unlike the frequentist approach, it incorporates both observed data and, when available, a priori information (called priors) to update the probability of a hypothesis as new data are collected.

CRO / A/B testing :

The Bayesian approach can be used to answer concrete questions such as :

"What is the probability that variation B is better than variation A?"

Rather than rejecting or rejecting a null hypothesis, it provides a direct, actionable probability, far more intuitive for business teams.

Advantages for the CRO :

• Provides results that can be used continuously, without waiting for a predefined sample size,
• Dynamically adapts to test evolution,
• Provides clear probabilistic estimates: for example, "there's a 92% chance that variation B is better than A",
• Facilitates rapid, progressive decision-making,
• Less exposed topeeking bias,
• Compatible with advanced customization or multi-variant testing contexts.

Tools using the Bayesian approach :

• Optimizely (Stats Engine)
• VWO (Bayesian Engine)
• AB Tasty (Bayesian option)
• Convert.com, Google Ads (Experiences), etc.

Points to watch :

• The choice and interpretation of priors can influence results if poorly calibrated,
• Less familiar to purely data/scientific teams used to traditional testing.

‍