Statistical significance
The point at which the gap between a variant and the control is too large to be explained by random chance alone.
Visitors behave differently from day to day, so two identical pages never show identical numbers. Significance tells you whether the difference you see is bigger than that normal noise.
Significance depends on sample size, so set the size before you start and let the test finish. Ending a test the first time it looks significant, or after a few days, is one of the most common ways to ship a change that does not actually work.
How Elevate calls a winner
Elevate uses a Bayesian model. A result is significant once the probability to win passes 90%, the recommended threshold, and the test has met its minimum visitors, conversions and running days.
Use the AOV A/B test calculator to see how many orders each variant needs before you launch.
Put it to the test on your store
Test prices, themes, pages, shipping and checkout on Shopify, then ship the winning variant.
