Build a dynamic segment

Overview

Dynamic segments recompute on a schedule or near-real-time stream, keeping marketing and support lists current without CSV uploads. Start from behaviors, then layer traits and predictive scores when available.

Steps

1. Create rules referencing recent events (ViewedProduct within 3 days) and stable traits (plan_tier).
2. Preview overlap with an existing holdout segment to avoid audience collisions.
3. Connect sync jobs to one primary channel first; expand after monitoring error budgets.

{
  "name": "intenders_no_purchase_7d",
  "logic": "AND",
  "clauses": [
    { "event": "ProductViewed", "window_hrs": 168, "min": 3 },
    { "event": "Purchase", "window_hrs": 168, "max": 0 }
  ]
}

4. Enable change-data capture exports if your ESP expects incremental diffs instead of full snapshots.

Troubleshooting

• Audience shrinkage: check whether upstream tagging broke—compare segment trend to ingest volume.
• Mismatched counts vs warehouse: confirm warehouse latency; segments often lead warehouse by an hour.
• Activation throttling: reduce sync frequency when destination APIs rate-limit.