h1
style
article-header-section

Federated Audience Composition lets you query your data warehouse directly from within Real-Time CDP — no data copying required. If you have rich customer data sitting in the warehouse that you've never been able to activate for marketing, this article is for you.

Why Federated Audience Composition

Getting offline data into Real-Time CDP in a quick, governed and temporary fashion used to be impossible. Marketing wants personalization. Tech wants lean data. Legal wants governance, and the leadership team needs it done yesterday.

This is the tension that Federated Audience Composition was built to resolve.

In our organization, data composability means that marketing teams can work with data where it already lives, rather than forcing a one-size-fits-all ingestion pattern for every use case. Not every attribute should live inside Real-Time CDP. Some data is sensitive. Some only exists to qualify a customer for a single campaign at a moment in time. A composable data strategy means having flexibility to decide when to ingest, when to federate, and when to simply query.

The challenges we faced prior

Data duplication and governance risk. Every time marketing needs a new audience that uses data that isn’t currently in Real-Time CDP, we need a way to ingest that data onto the profile. You can do this through writing profile attributes in a journey or simply setting up a new ingestion pattern. However, this can create redundant copies of data, raise concerns about compliance, and bloat the profile.

Marketing dependency on tech for every audience using data NOT on the profile. Building audiences from complex warehouse attributes required engineering tickets, SQL requests, and multi-week turnaround times. Marketers had visibility into what they wanted but couldn't self-serve and campaign velocity suffered.

The "all-or-nothing" ingestion problem. Our data strategy was binary: either data lived in Experience Platform (fully ingested, profiled, governed), or it didn't exist for marketing purposes at all. There was no middle ground for high-value, campaign-specific data.

Our use case: Personalized Direct Mail

The use case that crystallized the value of Federated Audience Composition for us was a personalized direct mail campaign. We needed to activate a targeted audience, built from temporary, campaign-specific attributes living in our data warehouse, and route that list downstream to a print vendor for a physical direct mail piece.

Here's what made this use case uniquely suited for Federated Audience Composition:

By combining Real-Time CDP with Federated Audience Composition, we were able to temporarily ingest the data, compose the audience, and activate it to the print vendor — all without a single row of underlying data being copied into Experience Platform, and all within a day. The result was a faster, cleaner, and more governed than any ingestion à activation pattern we had used in the past.

Getting started — building the foundation for data federation

Data federation isn't a workaround - it's a strategic architectural pattern. Platforms like Databricks, Snowflake, and Redshift are probably already the most authoritative sources of data in your organization. The question was never whether that data is valuable; it's whether your activation layer can reach it without creating a compliance headache or leaving a graveyard of stale attributes on the profile long after the campaign wrapped. That's the problem Federated Audience Composition solves.

Federated Audience Composition brings a zero-copy query pattern into Experience Platform. This approach doesn’t contribute to profile bloat, preserves the warehouse as the system of record, and still makes the resulting audience fully actionable within Real-Time CDP.

For companies applying this to their data strategy, the recommendation is to start with use cases where your warehouse holds data that is either (a) too sensitive to copy, (b) only needed temporarily for a specific campaign, or (c) too large and detailed to justify full ingestion.

How the Initiative Came Together: Co-led by Tech and Marketing.

Our Federated Audience Composition initiative was driven by a quick-turnaround marketing use case and enabled by tech.

Marketing defined the audience criteria and validated the data inside Databricks.

Our tech team ingested the data through Federated Audience Composition and made it temporarily available in Real-Time CDP.

Marketing then activated the campaign through a destination.

The collaboration worked because both teams respected their ownership boundaries. Tech controlled the data processes and marketing controlled validation and activation.

Warehouse Compatibility

For our implementation, we connected to Databricks.

Federated Audience Composition currently supports connections to the following platforms:

If your org runs Databricks, the connection setup is straightforward — your data engineering team will configure the cluster connection and service principal credentials, then define which schemas are accessible from the Experience Platform side. The key governance decision is which delta tables or views to expose, and at what column granularity.

Data Decisions — Knowing when to federate vs. ingest

One of the most common questions marketers and data architects face when configuring Real-Time CDP is: should this data be streamed, batched, or federated? The answer depends on three factors: latency requirements, data governance constraints, and frequency of reuse.

Here's how we think about it:

Use Streaming Ingestion when:
Use Batch Ingestion when:
Use Federation when:

Default alt

What Warehouse Data Makes Sense for Marketing?

Not all data is worth federating. In our case, the data we needed…

The Audience:

For our personalized direct mail campaign, here's how we constructed the federated audience:

Business objective: Reach customers who qualified for a targeted offer based on CRM and offline attributes living outside of Real-Time CDP, and route that list & attributes to a print vendor for a personalized printed mailer.

Audience logic:

The resulting audience was composed entirely through Federated Audience Composition's drag-and-drop canvas in Experience Platform, with the underlying query pushed down to Databricks. The audience was then activated to the vendor (through a Real-Time CDP destination).

Default alt

Best practices and lessons learned

Lessons from the Field

Implementing Federated Audience Composition requires alignment across marketing, data engineering, and governance teams. Here are the most important things we learned:

1. Start with a governed schema, not a wide-open table. When your data team exposes warehouse tables, resist the temptation to bring in every column. Work with your data team to define a limited, useful schema that exposes only the attributes marketing actually needs for audience qualification. This limits risk and makes the composition UI easier for marketers to navigate. Just because Federated Audience Composition gives you flexibility on broader data, doesn’t mean you’re allowed to get lazy and bring in everything.

2. Agree on data freshness SLAs before you build the audience. Federated queries run against your warehouse at composition time. If your warehouse data is refreshed nightly, your federated audience reflects last night's state.

3. Use it for qualification, not for the entire profile. Federated Audience Composition is most powerful when used to add a qualifying "layer" on top of known Experience Platform profiles, not to recreate the entire profile from scratch. Think of it as: Experience Platform owns the activation and your warehouse owns the campaign-specific attributes. Combine them in a composition for the best of both worlds.

4. In a Tech-led model, define a clear intake process for new federated schemas. Because our data engineering team owns the Databricks connection and schema governance, marketing can't simply add new tables at will. We established a lightweight intake process…

This prevents schema sprawl and keeps governance intact without creating a bottleneck.

5. On Databricks: use delta table views, not raw source tables. Views let the team pre-apply row-level filters, column masking, and business logic (like suppression flag resolution) before Federated Audience Composition issues a query.

How to create your first federated audience

Step 1: Establish the federated database connection

In Adobe Experience Platform, navigate to Federated Data > Federated Databases. Select Add Federated Database and choose your warehouse type. You'll need to provide your server hostname, HTTP path, and a personal access token or service principal credentials. Adobe will use this connection to push queries to Databricks at composition time. No data is pulled into Experience Platform at this step.

Step 2: Create and configure your federated schemas

Once the connection is established, navigate to Federated Schemas. Create a new schema mapped to the warehouse table(s) you want to use for audience composition. NOTE: Select only the fields marketing needs and avoid exposing sensitive columns that aren't required for audience logic or export. Publish the schema when ready.

Step 3: Data Model Creation

Navigate to Federated Data > Data model, click on "Create data model" and create a label for it. Select the schema that was created in step 2 (and any other schemas that may be linked to it) and click "Save". If additional schemas are to be selected as part of the data model, you can select them and then join the two schemas together on a common attribute or identity.

Step 4: Build your composition in the Audience Composition canvas

Navigate to Audiences > Create Audience > Compose Audience. In the composition canvas, add a Query activity and select your federated schema as the data source. Use the visual query builder to define your inclusion/exclusion logic — filters, attribute conditions, and value thresholds. You can layer in additional activities like Split (to divide the audience into sub-segments), Enrichment (to add additional attributes), and Exclusion (to remove suppressed profiles).

Step 5: Save and activate the federated audience

Once your composition is complete, save and publish the audience. The audience will appear in your Experience Platform Audiences inventory and can be activated to any destination.

Step 6: Monitor and refresh

Use the Audience portal to monitor audience membership and schedule audience refreshes aligned with your warehouse data freshness SLA. For campaign-specific audiences, you can archive the composition after the campaign ends.

Measuring success and optimizing your strategy

How We Measure Success

For our direct mail use case, success was measured across three areas:

  1. Audience precision: Did the federated audience qualify the right customers?  We validated our audience list and corresponding attributes against the source data (post-composition) to confirm that our export is what we expected.
  1. Campaign performance: For the direct mail campaign, our primary success metrics were response rate (offer redemption) and downstream conversion. Because Federated Audience Composition enabled us to use more precise qualifying attributes (attributes we couldn't have used via standard ingestion) our audience was more tightly targeted and personalized than prior campaigns.
  1. Operational efficiency:  Utilizing pre-built data reduces the need for global evaluations and multiple rounds of QA.  We were able to take 1-2 sprints worth of work down to a single day.

Federated Audience Composition paired with Real-Time CDP enabled a campaign that would not have been possible through any other ingestion or activation pattern we had available.  The combination of qualifying data that lived outside of Real-Time CDP, a strict print deadline, and a downstream offline activation requirement made Federated Audience Composition the only viable path within the timeline that we had.

Additional resources

Adobe documentation & learning

Adobe blog & thought leadership

style
article-content-section