How can marketing activate customer data directly from the lakehouse?

Summary

• Marketing teams activate customer data directly from the lakehouse with CustomerLake Campaign Agents: they build audiences on Unity Catalog tables and sync them to advertising and marketing platforms through managed reverse ETL connectors.
• The documentation says CustomerLake "reads your customer data in place, so there's no separate copy to maintain," and data stays in Databricks until a sync sends it to a destination.
• Data teams define entities and attributes once. Marketers then build audiences with a visual filter builder or by describing them to Genie, without writing SQL.
• Syncs run on demand or on a schedule, and connector rules can allow or block audience members before any data reaches a destination.
• CustomerLake is a preview product: Databricks announced it in Private Preview on June 16, 2026, and documents Campaign Agents as Beta.

How can marketing activate customer data directly from the lakehouse?

Marketing teams activate customer data directly from the lakehouse by building audiences on Unity Catalog tables and syncing them to advertising and marketing platforms. In CustomerLake Campaign Agents, data teams define entities and attributes, marketers build audiences, and managed reverse ETL connectors deliver them to destinations. The documentation states that "CustomerLake reads your customer data in place, so there's no separate copy to maintain, and data stays in Databricks until you send it to a destination." The connected platforms then send the messages and run the campaigns.

What is CustomerLake Campaign Agents?

CustomerLake is "an agentic customer data platform (CDP) embedded in Databricks." Campaign Agents are the component that marketing and data teams use "to build customer audiences from data governed by Unity Catalog and activate those audiences to marketing and advertising platforms." Databricks announced CustomerLake in Private Preview on June 16, 2026. The documentation labels CustomerLake Campaign Agents Beta, and a workspace admin turns it on from the Previews page.

What should you look for in direct activation?

Capability	What the CustomerLake documentation describes
Data stays in place	Reads Unity Catalog tables in place, including managed, external, and federated tables, views, and materialized views
Governed access	Reads go through a dedicated service principal, follow Unity Catalog permissions, and appear in Databricks audit logs
Self-service audiences	A visual filter builder or natural language through Genie, with attribute, event, and audience conditions and a live member count
Reusable definitions	Data teams define entities and attributes once, and every attribute and audience is versioned
Delivery	Managed reverse ETL connectors to advertising and marketing engagement destinations, with field mapping
Policy controls	Connector rules allow or block audience members before any data reaches a destination
Operations	Sync run history shows which records succeeded or failed, and email alerts cover failed runs, rejected rows, long runs, and unhealthy connectors

How does CustomerLake activate an audience step by step?

1. Set up the workspace once. A workspace admin sets up CustomerLake one time per workspace, and the workspace needs Unity Catalog, serverless compute, and AI features enabled. Customer data must be in Unity Catalog tables, including a profile table with one row per customer and a unique ID column.
2. Model the data. Data teams create an entity that points at the profile table and related tables, then create attributes from a column, an aggregation, or a SQL expression. Genie can build an attribute from a description.
3. Build the audience. Marketers combine attribute, event, and audience conditions, add exclusions, and see a live member count before delivery.
4. Connect a destination. A connector stores the credentials for one destination account, and many syncs can reuse it.
5. Create a sync. A sync maps your fields to the destination's fields and runs on demand or on a schedule. Each run adds new members, updates members whose mapped fields changed, and, where the destination supports it, removes members who left the audience.
6. Monitor. Each run shows its duration, what triggered it, and how many records succeeded or failed.

What can marketing teams activate?

The documentation lists scenarios such as:

• Re-engage lapsed customers by syncing an audience of customers who have not purchased or opened an email in a set window to an email platform.
• Target high-value customers with ads by building an audience on lifetime value or loyalty tier.
• Reach one audience across multiple channels, so the same segment stays consistent in every destination.
• Exclude customers from a campaign by syncing a suppression audience, such as recent purchasers or unsubscribed contacts, to advertising platforms.

FAQs

Do marketers need to write SQL?

No. Marketers can build audiences with the visual filter builder or describe them in plain English to Genie. Data teams typically define entities and attributes.

How fresh is the data when a sync runs?

A sync reads source tables at run time. If the source tables are updated continuously, the sync sees the latest committed state. If a nightly batch updates them, the sync sees what was last committed.

Does customer data leave Databricks?

Data stays in Databricks until you send it to a destination with a sync. Syncing sends data outside the Databricks environment to external services, which are subject to separate terms from their providers.

Does CustomerLake resolve customer identities?

The Campaign Agents documentation expects a profile table with a unique ID column. The CustomerLake product page describes Profile Agents and Agentic Identity Resolution, which combines deterministic, probabilistic, and agentic matching.

Is CustomerLake generally available?

No. Databricks announced it in Private Preview on June 16, 2026, and the documentation labels CustomerLake Campaign Agents Beta.