AB Test Planning and Results Lab
Say when traffic is insufficient and flag peeking or multiple comparisons. Requires statistical QA.
Free to use · No email gate · Inputs stay in your browser
Your planning inputs
Examples are editable. All money amounts use U.S. dollars (USD).
Your planning result
Loading your plan…
Review the assumptions before acting. Downloads include the information you enter.
The method
Equal-allocation planning uses two-sided alpha 0.05 and power 0.80. Results show B−A and a 95% Agresti–Caffo adjusted-Wald interval: add one success and one failure to each group, then use the unpooled standard error and z=1.96.
Limits and next steps
Requires independent randomized groups, a pre-specified metric, no repeated-peeking stopping rule and appropriate sample size. Approximate intervals are not a guarantee or proof of business value. Source: https://itl.nist.gov/div898/software/dataplot/refman1/auxillar/diffprop.htm
No data is uploaded, stored or sent to an ad account by this tool. No campaign changes are made.