How Is Confidence Calculated?
"Chance to Win" is the probability that your variation has to perform better than the control. The higher the "Chance to Win," the more certain you are that your variation will win against the control. If a variation within an experiment has a 99% chance to win, that reflects a high level of certainty that it will outperform the control. It is not enough to rely on high "chance to win" levels. To ensure that you are identifying a true winner against a control,
- Allow your experiment to run a minimum of one week (preferably two weeks)
- Make sure that a winning variation and the original have each recorded 100 conversions.