Statistical Power Calculator

Statistical power is the probability that a test correctly detects a real difference between two groups, given that a real difference of the size you specified actually exists. An 80% power means the test would catch that real difference in roughly 4 out of 5 identical experiments; the remaining 1 in 5 would come back as a false negative purely from sample noise.

Because a non-significant result only tells you something useful if the test had enough power to detect the effect you cared about in the first place. Calculating power after the fact, using the sample size the test actually got, tells you whether "no significant difference" means the variant genuinely did not help, or the test simply was not big enough to know either way.

80% is the most common threshold used in conversion rate testing, meaning you accept roughly a 20% chance of missing a real effect of the size you specified. Some teams use 90% for higher-stakes decisions, which requires a noticeably larger sample size to achieve.

A significance calculator tells you whether the difference you already observed in your data is likely real or likely noise. A power calculator tells you, before or after the fact, how good your sample size is at detecting a difference of a given size at all. You can have a significant result with plenty of power, or a non-significant result with too little power to have found anything reliably.

Extremely high power usually means you are running far more visitors through a test than necessary, which costs time and traffic without much added benefit once you are comfortably above 80 to 90%. It is not harmful the way low power is, but it is often a sign the test could have reached a decision sooner with a smaller sample.

No. This uses the standard two-proportion formula for a single variant against a single baseline. Testing three or more variations at once changes the math because of the added comparisons, so treat each pairwise comparison separately if you are running a multi-variant test.