You're evaluating an experiment on a collaborative product where one user's treatment can affect teammates through shared workspaces, projects, and tasks. The team is trying to measure the impact of inviting teammates into a workspace without contaminating the control group.
How do you test a feature that has strong network effects, like inviting teammates to a workspace?
Choosing the right unit of randomization under spilloversReasoning about causal validity when SUTVA failsQuantifying the power loss from cluster randomizationPicking an analysis method consistent with the experiment design