Minimum detectable effect (MDE) is the smallest estimated effect size that an experiment is designed/powered to detect. Small effects are harder to pick up, and selecting an MDE allows us to estimate how many users and time an experiment will take.

• There are some experiments where we can know up front, will never succeed (we don’t have enough users or time).
• There are other experiments that we can be confident will end early (if there is an effect).
• This helps prioritize and sets the goals of the experiment.