For the complete documentation index, see llms.txt. This page is also available as Markdown.

Impact analyses

Build an impact analysis in Hypertune to see the impact of an experiment on multiple metrics in a single view.

Choose between:

  • Frequentist analysis — shows you conversion rate uplift with confidence intervals

  • Bayesian analysis — shows you the probability of each variant being the best

Both methods automatically apply adjustments for:

  • Sequential testing — avoids the peeking problem, enabling you to view results at any time and ship a variant if it has a significant result

  • Multiple comparisons — adjusts for the family-wise error rate, enabling you to compare more than two variants (in a multi-arm test), to determine an overall winner

Last updated