Audience Segmentation and Destinations/Activation
Segmentation fundamentals
Audiences (segments) are defined by Segment Builder using PQL (Profile Query Language) or the visual builder. Key concepts: streaming segmentation (evaluates events near-real-time) versus batch segmentation (scheduled). Edge segmentation evaluates on-device for low-latency use cases. Questions test when each applies: streaming for event-driven membership, batch for large historical sets.
Evaluation and refresh
Understand segment evaluation methods and refresh cadence. A trap: assuming a newly created streaming segment includes historical data - it typically only captures members from creation forward unless backfilled. Another trap: qualifying events vs. profile attributes - event-based criteria require the event to be profile-enabled and stitched.
Destinations and activation
Destinations connect audiences to activation channels. Know the difference between profile export destinations (file/streaming to cloud storage, ad platforms) and edge/personalization destinations. Activation maps audience members and attributes to the destination's expected schema. Destination flow concepts: source (audience), destination, mapping, and schedule. Traps include expecting real-time activation from a batch destination, and mismatched identity mapping (destination expects a namespace not present in the audience).
Governance at activation
Data usage labels and marketing actions on destinations enforce policy - a destination with a marketing action that conflicts with a label blocks activation. Questions often hide this: audience looks fine but export fails due to policy.
Rule of thumb
For any activation question, check three gates: is the audience populated (evaluation method), is the identity mapping valid for the destination, and does governance permit the export?
Worked example: A streaming audience exports to an ad platform but returns empty. The destination is batch-scheduled and the audience's identity namespace isn't mapped. Fix mapping and confirm schedule; governance is fine if no label conflict.