The Segment That Never Reaches the Destination
A travel company built a 'Cart Abandoners - 7 Day' streaming audience that qualifies on an abandonment event and should exit when a purchase event arrives. The audience is activated to an ad platform destination that expects a daily batch of hashed emails. In testing, the audience count in the UI looks correct, but the destination receives roughly 40% fewer profiles than expected, and users who purchased yesterday are still appearing in today's export. The destination's own match rate is low. The team confirmed the destination is connected and the dataset is profile-enabled. The audience uses a streaming evaluation, and the destination is configured with a daily schedule.
1. Account for the gap between UI count and delivered profiles, and for the stale purchasers.
Read the worked answer
Two separate mechanisms explain the two symptoms.
Delivered < UI count (40% gap) The destination expects hashed emails. A profile qualifies for the audience but can only be exported if it carries an identity the destination accepts and that identity is present and properly formatted. Profiles that qualified via an anonymous or cookie identity, or whose email is missing/unhashed, are counted in the UI but silently dropped at export. Low destination match rate corroborates this - the emails that did arrive may be hashed with a different algorithm or normalization than the destination expects.
Stale purchasers The audience is streaming, but the destination runs on a daily batch schedule. Exit events (purchase) update the profile in near real time, but the export reflects audience membership at export time only if the exit is processed before the export snapshot. If exits are evaluated on a different cadence than the export, or if the purchase event does not actually carry the identity used for audience evaluation, the exit never takes effect. A common cause: the purchase event lacks the email identity, so the exit applies to a different (anonymous) profile than the one in the audience.
So: the UI counts qualified profiles; the destination receives exportable profiles. Those are different populations.
2. Recommend concrete fixes for both symptoms, in priority order.
Read the worked answer
1. Fix identity continuity for exits (highest priority - it's a correctness bug)
- Ensure the purchase event carries the same identity (email) used to qualify the abandonment audience, so the exit lands on the same profile.
- Verify the exit condition is actually configured on the audience and that streaming evaluation processes it.
2. Align export cadence with audience semantics
- For a 7-day cart-abandoner window with same-day purchase exits, a daily export is acceptable only if the export snapshot runs after exit processing. Confirm ordering, or move to a more frequent export if the destination supports it.
3. Fix the exportability gap
- Audit which identity the destination mapping uses and confirm every qualifying profile is expected to have it. For profiles without email, either accept they are non-exportable (and adjust expectations) or enrich them via a CRM join before activation.
4. Fix hashing/normalization
- Match the destination's required hash (algorithm, lowercasing, trimming) exactly; a mismatch explains low match rate even when emails are present.
5. Instrument it
- Track qualified vs. exported counts per run and alert on divergence, so this class of gap is visible before it erodes campaign performance.