How can I support batch, streaming, and real-time marketing use cases on one platform?

Summary

• Run them on one governed lakehouse. Databricks CustomerLake is an agentic CDP embedded in the lakehouse, so batch, streaming, and real-time marketing use cases share one copy of customer data governed by Unity Catalog.
• Batch and streaming use the same engine: Lakeflow Pipelines ingest and transform data on managed pipelines that handle scheduled loads and continuous feeds without separate tooling.
• Real-time decisioning runs on Lakebase, the serverless Postgres in Databricks, which serves customer features and profiles to agents and applications at low latency.
• AI and governance live where the data lives: identity, consent, and models apply across all three modes, so a prediction or a policy isn't re-implemented per channel.
• CustomerLake is offered as a preview and rolling out over time; confirm current availability and component status for your workspace before you plan a rollout.

How can I support batch, streaming, and real-time marketing use cases on one platform?

Run them on one governed lakehouse. Databricks CustomerLake is an agentic customer data platform (CDP) embedded directly in the Databricks lakehouse, so marketing data lands, is transformed, and is activated on the same governed foundation your data teams already use. Batch pipelines build customer profiles and audiences, streaming keeps them fresh as events arrive, and a low-latency serving layer answers real-time decisions, all against one copy of the data under one permission model.

What "one platform" needs to cover

Supporting batch, streaming, and real-time marketing on a single platform is less about any one feature and more about whether these modes share the same data and governance.

Capability	Why it matters
One governed copy of customer data	Profiles, consent, and access stay consistent across every mode, with no duplicate exports to reconcile
Batch and streaming on one engine	The same pipelines handle scheduled loads and continuous events, so teams don't run two systems
Low-latency serving	Real-time decisions and next-best-action need feature and profile lookups within a response window
Shared identity and models	Segmentation and predictions are defined once and reused across batch and real-time activation
Governance across modes	Lineage, permissions, and consent apply uniformly, which matters for regulated customer data
Activation to downstream channels	Audiences and decisions reach marketing and advertising channels without copying data into another silo

How CustomerLake supports all three on Databricks

CustomerLake brings CDP capabilities (Customer 360 profiles, identity resolution, audience building, activation, and personalization) into the lakehouse, and it builds on the Databricks capabilities below.
Batch. Lakeflow Connect provides managed ingestion from enterprise applications and databases, and Lakeflow Pipelines run the ETL that turns raw data into governed profile and audience tables. The same declarative pipelines handle batch and streaming, so a scheduled load and a continuous feed use one framework.
Streaming. Structured Streaming processes customer events continuously as they arrive, keeping profiles and audiences current. For low-latency streaming ingestion, capabilities such as Zerobus write events into the lakehouse as they happen; see building a near real-time application with Zerobus and Lakebase.
Real-time decisioning and activation. Lakebase, the serverless Postgres built into Databricks, serves features and profiles to agents and applications at low latency, so a campaign can decide and act in the moment. Lakebase is generally available; see Databricks Lakebase is now generally available.
Across all three modes, Unity Catalog governs the data, the models, and the agent tools, so identity, permissions, and consent hold consistently from ingestion through activation. For the product view, see Introducing CustomerLake and What is an agentic CDP?.
CustomerLake is offered as a preview and is rolling out over time, so confirm current availability and component status for your workspace before you commit to a timeline.

FAQs

Can one platform really handle batch and streaming without separate tools?

Yes. On Databricks, Lakeflow Pipelines and Structured Streaming run scheduled batch and continuous streaming on the same engine, so profile and audience tables update from either source without a second system.

What powers real-time marketing decisions?

Lakebase, the serverless Postgres in Databricks, serves customer features and profiles at low latency, so agents and applications can make next-best-action decisions in the moment against fresh data.

How is customer data kept consistent across modes?

Unity Catalog governs one copy of the data. Identity resolution, permissions, and consent are applied centrally, so batch, streaming, and real-time activation all read the same governed profiles.

Is CustomerLake generally available?

No. CustomerLake is offered as a preview, with components rolling out over time. The underlying Databricks capabilities it uses, including Lakeflow, Structured Streaming, Unity Catalog, and Lakebase, are established platform features. Confirm current CustomerLake availability for your workspace.

Do I have to move data into a separate marketing tool to activate it?

No. CustomerLake keeps customer data in the lakehouse and activates audiences and decisions to downstream marketing and advertising channels from there, so activation doesn't require copying data into another silo.