Audiences, Targeting, and Personalization Strategy
Target audiences are built from rules: custom (mbox/params, profile attributes, geo, browser, date), Analytics segments (shared via A4T), Audience Manager segments, AEP segments, and Target's built-in audience library. Understand the difference between an audience (who) and an offer/experience (what). Profile attributes persist in the visitor profile; mbox parameters are page-scoped and must be passed at request time.
Know the targeting hierarchy: activity-level audience, experience-level audience (XT), and offer-level targeting. For Recommendations, inclusion rules filter the item catalog; for AP, targeting can restrict who enters the model.
Traps: assuming an Analytics segment updates in real time (it may lag); confusing profile parameters (persisted) with mbox parameters (transient); forgetting that audience membership can be evaluated at the edge vs on-device with different data availability. Also watch for "audience overlap" questions - a visitor can qualify for multiple activities, and collision rules/prioritization matter.
Personalization strategy: segment by behavior and value, not just demographics; use first-party data; keep audiences small enough to be actionable but large enough to be measurable; always include a holdout/control to prove lift.
Rule of thumb: if a segment can't reach statistical power in the test window, broaden it or extend the window - don't launch a test you can't read.
Worked example: target high-value cart abandoners. Build a profile attribute loyaltyTier=gold passed at login, plus an event-based audience for cart-abandon in the last 24h, then XT to show a recovery offer with a control group.