Peeking
Checking a test while it runs and stopping it as soon as one version looks ahead, which inflates the chance of a false win.
Early results swing a lot. If you check every day and stop the first time a variant crosses the significance line, you will declare far more winners than really exist.
The fix is simple: set the sample size and duration before launch, and make the decision only when the test reaches them. Watching the numbers is fine. Acting on them early is the problem. Elevate holds back a winner until the test meets its minimum visitors, conversions and running days.
Put it to the test on your store
Test prices, themes, pages, shipping and checkout on Shopify, then ship the winning variant.
