Experiment Stats

Compute real inferential statistics for A/B and multivariate marketing experiments, then apply a dual promote-gate to decide whether a variant actually won. The p-value and confidence interval behind that call come from bootstrap resampling and a hand-implemented Mann-Whitney rank-sum test, run directly against the outcome data you provide.

What it does

Given a control group and one or more variant groups of outcome data, scripts/ab_stats.py: