POC Readout II
3 min
Content
Generated by AI. Be sure to check for accuracy.Meeting notes:
- Temporary Table Persistence in Federated Audience Composition: Sheyla and Andres clarified that temporary table names persist in the schema list due to transitions between activities and scheduled compositions, as confirmed by engineering, addressing an outstanding question from the previous meeting.
- Table Name Persistence: Temporary table names appear in the schema list because they are saved during transitions between activities or nodes in Federated Audience Composition, and they persist when the composition status is not stopped or paused.
- Audience Creation and Enrichment Workflows: Sheyla led a walkthrough of the creation and enrichment of audiences 3 and 3A, detailing the use of propensity scores, segment membership, opt-in status, and experience events, with Alexandra Petrus providing input on operator usage and segment logic.
- Audience 3 Definition: Audience 3 was defined by qualifying users with a medium offline modeling or propensity score from the data lake and segment membership as past passengers (PPAX), with enrichment using the rank field and creation in AEP.
- Query Logic and Scheduling: The query for Audience 3 used rank operators to select profiles between 17 and 32, and the scheduler activity was configured to refresh the audience daily, specifying frequency, time, and validity period.
- Enrichment with Existing Attributes: The save audience activity enabled exposure of the composition in the audience portal, where further enrichment was performed using existing AEP attributes such as PPAX segment membership and medium propensity.
- Audience 3A Enrichment: Audience 3A built upon Audience 3 by adding criteria for web form submission and opt-in status, using the unsubscribe date field to identify opt-in users and enriching with experience events evaluated on the edge for real-time personalization.
- Operator Usage Clarification: Alexandra Petrus asked about using 'exists' for PPAX segment membership, and Sheyla confirmed that operators can be adjusted for inclusion or exclusion, with the segment logic working as expected for qualifying profiles.
- Profile Enrichment and Attribute Persistence: Sheyla explained the process of profile enrichment, highlighting the difference between federating data for audience composition and persisting federated attributes in AEP profiles using the save profiles activity, with Andres inquiring about attribute mapping and transformations.
- Profile Enrichment Workflow: Profile enrichment allows federated data, such as Shorex recommendations or propensity scores, to be persisted in AEP profiles by defining attributes in an XDM schema and mapping them from the federated source to the schema using the save profiles activity.
- Attribute Mapping and Schema Setup: Sheyla demonstrated creating a dedicated XDM schema and dataset for profile enrichment, mapping fields like booking ID, e-mail, product line, and sale date from the federated source to the schema object for storage in AEP profiles.
- Transformation Capabilities: Andres asked about transformations during attribute mapping, and Sheyla clarified that transformations must be performed before the save profiles step, utilizing available targeting activities such as enrich, reconcile, and change dimension within the composition canvas.
- Retention and Update Options: Federated audience composition retains qualifying profiles and segment membership for 30 days unless a daily frequency is set; with profile enrichment, users can select full refresh or incremental updates for persisted attributes.
- Governance, Privacy, and Data Access Considerations: Sheyla provided a comprehensive overview of governance, privacy, and data access controls for federated audience composition, addressing previous questions from Andres and confirming the applicability of privacy service, consent policies, data lifecycle, access labeling, and audit logs.
- Privacy and Consent Management: Privacy service handles consumer data access and deletion requests in compliance with global regulations, and consent policies can be enforced for automated downstream activation if configured.
- Data Lifecycle and Hygiene: Data lifecycle management and hygiene APIs allow for deletion of records and datasets, with these processes remaining applicable to federated audience composition.
- Access Labeling and Audit Logs: Governance policies and access labels can be applied at the audience, dataset, and composition level, with audit logs available for create, read, update, and delete operations in the Federated Audience Composition UI.
- Role-Based Access Control: Two new roles are automatically created for federated audience composition in the permissions UI, enabling assignment and management of user roles, with custom roles available for specific teams.
- Data Localization and Guardrails: Database connections must match the sandbox region, and asset limitations include up to 4000 active audiences per sandbox and 15,500 active streaming audiences for edge evaluation, with export file size recommendations for optimal performance.
- Benefits and Cost Efficiency of Federated Audience Composition: Sheyla summarized the benefits demonstrated in the POC, including reduced data movement, lower storage and processing costs, faster activation, compliance with sensitive data requirements, and licensing efficiency, with validation of daily scheduled federated audiences and enrichment strategies.
- Reduced Data Movement: Federated audience composition eliminates the need to ingest large datasets into AEP, reducing storage and processing costs and enabling validation of attribute reduction in production sandboxes.
- Faster Activation and Lower Latency: Audiences can be exposed and activated downstream more quickly without ETL onboarding, leveraging federated data for rapid refresh and activation, especially for daily scheduled audiences.
- Compliance and Sensitive Data Handling: Federated audience composition allows sensitive data to remain in the data warehouse for compliance, federating only necessary attributes into AEP profiles.
- Licensing and Cost Efficiency: Federating data from Databricks and managing retention strategies can reduce licensing usage and costs over time, with flexibility in deciding what to ingest versus federate.
- Audience Enrichment and Activation Workflows: Sheyla demonstrated advanced audience enrichment workflows, including enriching existing AEP audiences with federated data via the Federated Audience Composition destination, and explained activation processes for federated audiences and enrichment attributes, addressing questions from Brandon Farraye.
- Enriching AEP Audiences with Federated Data: Existing AEP audiences can be mapped to the Federated Audience Composition destination, exposed in the composition canvas, and further enriched with federated data using mapping and scheduling options.
- Activation and Attribute Mapping: Federated audience mappings, including enrichment attributes, are carried downstream for activation workflows, allowing personalization in destinations such as S3 and AJO, even if attributes are not persisted in the AEP profile.
- Personalization in AJO: Brandon Farraye asked about using enrichment attributes for personalization, and Sheyla confirmed that helper functions and expression editors in AJO can manipulate federated attributes for dynamic content, including case transformations.
- Integration with Orchestrated Campaigns: Brandon Farraye inquired about activating federated audiences in orchestrated campaigns; Sheyla explained that direct integration with Databricks-federated audiences will be available early next year, while traditional AJO campaigns currently support federated data.
- Provisioning and Sandbox Enablement for Federated Audience Composition: Andres, Anja, and Sheyla discussed the process for enabling Federated Audience Composition across multiple sandboxes, confirming that provisioning requests can extend access beyond the default two sandboxes, with considerations for database connections and licensing.
- Enablement Process: Federated Audience Composition was enabled for one sandbox during the POC, but can be provisioned for staging and production environments upon request, with licensing considerations for more than two sandboxes.
- Database Connection Strategy: Andres outlined a strategy to connect development and staging sandboxes to the development Databricks environment, and production sandbox to the production Databricks environment, to avoid conflicts.
- Licensing and Usage Limits: Anja confirmed that FAC can be used across all currently licensed sandboxes up to 500 gigabytes, with contractual obligations for multi-sandbox enablement.
- POC Wrap-Up and Next Steps: Sheyla, Alexandra Petrus, and Andres agreed to keep POC assets available until the end of the month for further validation and access, with Sheyla committing to upload recordings and provide follow-up resources before decommissioning.
- POC Asset Decommissioning: The team will wait until the end of the month before decommissioning POC assets to allow additional time for team members to access and validate the work.
- Resource Sharing: Sheyla will upload meeting recordings and link relevant Experience League pages to the landing page for reference and follow-up.
Follow-up tasks:
- Federated Audience Composition Enablement Across Sandboxes: Confirm whether Federated Audience Composition (FAC) can be enabled across all three sandboxes (development, staging, production) and communicate the contractual obligations regarding this. (Anja)
- Federated Audience Composition Provisioning: Initiate the process to enable FAC for the staging and production sandboxes as requested. (Sheyla)
- FAC Licensing and Sandbox Limitations: Verify if there are any licensing limitations for enabling FAC across more than two sandboxes and inform the team accordingly. (Anja)
- POC Asset Decommissioning Timeline: Wait until the end of the month before decommissioning all POC assets to allow team members additional time for access and validation. (Sheyla)
- FAC Enablement Resource Coordination: Connect the team with the appropriate resources to facilitate FAC provisioning on all licensed sandboxes up to 500 gigabytes. (Sheyla)