Netflix A/B exposure grain
Choose an exposure fact over stamping the experiment variant onto the viewing fact.
Concept
fact-table-grain
The primary modeling idea this problem reinforces.
Requirements
0
Business needs the model must satisfy.
An experimentation team needs engagement sliced by A/B variant. One proposal adds a variant column to the viewing fact; another models a separate exposure fact (one row per viewer-experiment-variant exposure).
Stamping the variant onto the viewing fact assumes one experiment per view and pollutes its grain. A dedicated exposure fact keeps experiment assignment at its own grain and supports many concurrent experiments.
Try the question first.
The discussion has other people's approaches and solutions. Give it a real attempt before you read them.