The series
Topics
Each topic takes one idea from experimentation and makes it something you can watch. Pick one to dive in.
-
The Eligibility Trap
A one-sided eligibility check quietly breaks the fairness of an A/B test. See how it corrupts the comparison — and the symmetric fix that restores it.
-
Don’t Condition on a Collider
An upsell lifts revenue per buyer, yet revenue per visitor falls. Watch how filtering on a collider can mislead you.
-
Significant, but Below the MDE
A fixed-sample test comes back significant, yet the observed lift is smaller than the MDE you designed for. That’s the normal case, not a broken test — see why half the wins land there.
Not-so-visual topics
Deeper reads that lean on math and prose more than animation — the ideas that don’t reduce to a single moving picture.