<b>If you didn't set your MDE before the test, your "significant" result is unfalsifiable.</b>

<b>If you didn't set your MDE before the test, your "significant" result is unfalsifiable.</b>
MDE (minimum detectable effect) is the smallest uplift your sample can reliably catch. Set it first; it determines sample size. Computing it after the fact lets you rationalize any outcome.

— Powered for a 5% lift but saw 2%? You can't conclude "no effect" — you were blind to it.
— Underpowered tests don't fail loudly; they whisper false nulls.

Read the number, not the story. [power=0.8 assumed unless stated]