A/B Test Calculator
Test equal conversion rates using two independent groups.
Formula
p = (xA+xB)/(nA+nB); z = (xA/nA−xB/nB)/√[p(1−p)(1/nA+1/nB)].
Conversion rate A ≈ 100×(100)÷(1,000) · Conversion rate B ≈ 100×(120)÷(1,000) · Z statistic A minus B ≈ ((100)÷(1,000)-(120)÷(1,000))÷sqrt((0.11)×(1-(0.11))×(1÷(1,000)+1÷(1,000)))
How to use
- Enter visitors and conversions for each independent group.
- Each group needs at least 10 conversions and 10 nonconversions.
- Reject equal rates at 5% when |z| exceeds the rounded cutoff 1.96.
Examples
- 60/100 versus 40/100 gives z ≈ 2.828427, beyond the 5% cutoff.
Frequently asked questions
Why reject small counts?
The normal approximation needs sufficient successes and failures; sparse data need another test.
Does significance mean B is better?
No. Negative z favors a higher observed B rate, but practical importance also depends on effect size.
Can I check repeatedly until significant?
Repeated peeking changes the error rate. Fix the analysis plan and avoid dependent observations.
Sources
Updated: