How can I unify customer data without creating another data silo?

Summary

• To unify customer data without a new silo, build profiles, audiences and activation on the governed platform where the data already lives, so no second copy needs its own security, governance and integrations.
• The CDP Institute separates packaged CDPs, which build profiles inside one system, from composable CDPs, which build them in your data warehouse from several components. Databricks describes a third design, an embedded CDP.
• CustomerLake is the Databricks Agentic CDP, embedded natively in Databricks and governed by Unity Catalog. Databricks says it does not move or copy customer data.
• Lakehouse Federation reaches customer data in other systems where it resides, and native managed reverse ETL connectors sync audiences and signals out to marketing tools.
• CustomerLake is in Private Preview. Interested teams contact their Databricks account team.

How can I unify customer data without creating another data silo?

Build the unified customer view on the governed data platform you already run, and bring customer data platform (CDP) functions to that data. A silo appears when profiles live in a second system that holds its own copy of the data, with its own security, governance and integrations to maintain. On Databricks, CustomerLake is built for the in-place approach. It is an Agentic CDP embedded natively in Databricks, so identity resolution, audiences and activation run where customer data already lives, and Unity Catalog governs access. CustomerLake is in Private Preview.

What is a data silo?

Databricks' guide to data silos describes them as data that teams collect, manage and store separately from one another, with access limited to certain groups. It says silos impede visibility and access, add inefficiency and cost, hinder governance and leave duplicate data that is hard to reconcile. A CDP turns into a silo when it keeps its own copy of customer data behind its own access controls.

How do CDP architectures differ on where customer data lives?

The CDP Institute describes two established designs. Databricks describes a third, an embedded CDP.

Architecture	Where profiles are built	Trade-offs reported
Packaged CDP	Within a single system that includes the tools to collect, assemble, store and access customer data	CDP Institute: faster deployment, less technical support and less ongoing maintenance, but buyers may pay for features they don't use and replacing a whole CDP is harder
Composable CDP	In a data warehouse, using several components for the different functions	CDP Institute: buyers reuse existing resources and buy only the functions they need, but IT needs more expertise and spends more time integrating and maintaining the components
Embedded CDP (CustomerLake)	Inside Databricks, on the governed data already there	Databricks: security, lineage and access controls are inherited from the Databricks Data + AI Platform and Unity Catalog

In the CustomerLake FAQ, Databricks adds that bundled CDPs require copying customer data into a separate proprietary system, creating another silo to secure, govern and integrate. It says composable CDPs reduce data movement but add another platform to manage, secure, govern and integrate. That is vendor positioning, so test it against your own requirements.

How does CustomerLake avoid creating a silo?

The CustomerLake FAQ answers the copy question directly: it does not move or copy customer data, because it is embedded in Databricks and runs CDP capabilities where the data already lives. The documented design points are:

• Governance stays in Unity Catalog. CustomerLake is governed through Databricks and Unity Catalog, which gives data and IT teams centralized control over access, lineage and auditability across customer data, audiences and AI assets. Unity Catalog is the governance layer for data and AI built into Databricks.
• Existing Databricks tools do the work. Databricks lists Unity Catalog for governance, Lakeflow for ingestion, AI and ML models, and Genie for natural language insights.
• Data in other systems stays reachable in place. Through Lakehouse Federation, teams can access customer data where it resides, whether in Databricks, other warehouses, cloud object storage or operational databases. Lakehouse Federation gives governed, read-only access to external data through Unity Catalog foreign catalogs.
• Identity resolution runs on the platform. Profile Agents prepare data, identify quality issues and support third-party enrichment to unify disconnected records into golden profiles. Agentic Identity Resolution combines deterministic, probabilistic and agentic matching, with LLM and human review for edge cases.
• Activation sends audiences out. Campaign Agents build audiences, recommend next-best actions and activate across channels. Native managed reverse ETL connectors sync audiences and signals from Databricks to martech, adtech and customer experience tools, which Databricks says avoids custom pipelines, separate activation vendors and additional data copies.
• Marketers work in natural language. An agent-first workspace powered by Genie lets marketers build audiences and ask questions without writing SQL, while data and IT teams keep managing data, pipelines, models and governance in Databricks.

What should I check before choosing an approach?

1. Where are profiles stored, and is that a second copy of your data?
2. Whose access controls, lineage and audit trail apply to the profiles?
3. Can data and IT teams reach the same data with the tools they already use?
4. How do audiences reach downstream tools, and does each hop create another copy?

FAQs

Does CustomerLake move or copy my customer data?

Databricks says no. CustomerLake is embedded directly in Databricks, so CDP capabilities run where your customer data already lives.

Does CustomerLake replace my existing martech tools?

Not necessarily. Databricks says CustomerLake is designed to work with the marketing, advertising, analytics and customer experience tools teams already use.

Can marketers use CustomerLake without writing SQL?

Yes, according to Databricks. The agent-first workspace is powered by Genie, so marketers can automate campaigns, ask questions, surface insights and build audiences in natural language.

Is CustomerLake generally available?

No. It is in Private Preview, announced on June 16, 2026. Databricks directs interested teams to their Databricks account team.