A conversion is a visitor completing a desired action, such as becoming a customer. The conversion rate is the proportion of visitors who convert: conversions divided by visitors. For example, 63 conversions among 500 visitors gives an observed rate of 12.6%. section 11.2
A measured rate is evidence, not certainty about the underlying rate. With a baseline of 10%, 63 conversions out of 500 visitors can be evaluated using a binomial distribution to ask whether the new design’s true rate is higher than 10%. section 11.2
Conversion rates support several practical questions:
- Estimate the probability that the underlying rate exceeds a threshold such as 0.1, using the area of a posterior beta distribution. section 11.10
- Check whether a low rate falls below a business-critical threshold such as 0.05. section 11.10
- Translate possible rates into decisions by assigning different utilities to launching or rejecting a design. section 12.3
from scipy.stats import beta 1 - beta(a, b).cdf(0.1)
Compute the posterior probability that the conversion rate exceeds 0.1 with SciPy’s beta cumulative distribution function. section 11.10





