AEP batch succeeded, but the profile and audience are stale: follow the record

A data team sees a green batch run and an updated row in the dataset preview. The marketer still sees yesterday's customer attributes or audience membership. A practitioner question in Experience League describes this exact split. The batch result answers whether data landed; it does not identify the last successful step for one customer.

Start with one known identity

Choose a test record with a distinctive changed attribute and record its batch ID, dataset ID, identity namespace, identity value and ingestion time. In AEP Monitoring, compare data-lake ingestion, identity and Profile processing for that batch. Adobe's dataset guide says a Profile-enabled dataset can expose counts of ingested identity and profile records. An empty or failed Profile stage calls for a different fix than a successful Profile update hidden by merge behavior.

Compare the Profile actually used

Look up the record under the identity the audience uses. Check whether the new data formed a fragment for another identity, then compare the active merge policy and source precedence. A value can be present in a dataset while another fragment wins in the assembled Profile. Record the winning fragment before changing policy; a global policy change can affect many audiences.

Check the audience clock last

Confirm the audience's evaluation method and last completed job. Adobe distinguishes batch, streaming and edge evaluation. A batch audience can be correct for its last snapshot while new Profile data awaits the next evaluation. Test a controlled evaluation where supported, then check membership for the same identity. Escalate to Adobe with the batch, profile and audience job IDs if the stages disagree.

Adovo’s view

Green is a property of a step, not a customer journey. The useful unit of proof is one identity moving from ingestion to assembled Profile to audience decision. When each boundary has an owner and timestamp, stale activation stops being a blame game.