Calculating metrics using the Bayesian method

To calculate conversions to goals, Varioqub employs the Bayesian method. The Bayesian method helps calculate the probability of the rate of user conversion to a certain goal being higher for the test group rather than for the control group.

The probability is calculated by the following formula:

$P(\lambda_T > \lambda_C​ ​) = sum_{i=0}^{\alpha_T - 1} \exp(logBeta(\alpha_C + i,\beta_C + \beta_T) - \log(\beta_T + i) - logBeta(1 + i, \beta_T) - logBeta(\alpha_C, \beta_C) )$
where $logBeta(x,y) = logGamma(x) + logGamma(y) - logGamma(x+y)$.

The probability is displayed in the experiment results report: Probability test > control. If the calculation shows that the test variant is better than the control one, it's highlighted in green. If it's worse, it's highlighted in red.