Databricks

What it is. Lakehouse platform combining Spark-native compute, Delta Lake storage, and integrated ML/AI tooling.

Strengths in CDP context.

• Native fit for heavy ML/feature-engineering workloads.
• Delta Lake handles streaming and batch in a unified table format.
• Strong unstructured-data support (ideal for clickstream, mobile event streams).

Native reverse ETL via Lakebase (2026). Databricks Lakebase is a fully managed Postgres-compatible database integrated with Unity Catalog. It enables activation of gold-layer Databricks tables directly into operational applications (CRMs, marketing platforms, APIs) via Postgres wire protocol — without requiring a third-party reverse ETL tool such as Hightouch or Census. Lakebase removes an architectural layer from Databricks-centric composable CDP stacks and eliminates the operational overhead of managing a separate activation tool.

Performance claim (vendor-asserted, unverified by third-party benchmark): Databricks states sub-10ms reverse ETL latency from gold-layer table to operational application. This claim is sourced from a Databricks marketing blog post and has not been validated by independent benchmarks. Apply with the vendor-asserted qualifier until a third-party benchmark source is available.

Competitive note: Lakebase positions Databricks as an alternative to standalone reverse ETL platforms (vendor.hightouch, vendor.census) for organizations whose data backbone is already Databricks-native. For Snowflake-primary or BigQuery-primary organizations, Hightouch and Census remain the natural activation layer.

CustomerLake — agentic CDP (2026, Private Preview). Databricks announced CustomerLake, a full-stack agentic Customer Data Platform built natively into the Databricks lakehouse. This repositions Databricks from CDW+reverse-ETL (Lakebase) to a composable-to-packaged bridge: organizations that already run analytics workloads in Databricks can build an agentic CDP layer without adopting a separate packaged CDP.

Key capabilities (all vendor-asserted, Private Preview — apply qualifier until confirmed by independent benchmark or analyst source):

• Agentic Identity Resolution (AIR): Combines deterministic, probabilistic, and AI-driven identity matching within the Unity Catalog lakehouse. No external MDM required for Databricks-native deployments.
• Profile Agents: Automate transformation of raw event and behavioral data into business-ready unified profiles. Analogous to identity-stitching in packaged CDPs (e.g., vendor.salesforce-data-360, vendor.adobe-experience-platform) but lakehouse-native.
• Campaign Agents: Build audiences, recommend next-best actions, and activate across channels. Covers the audience-segmentation-to-activation workflow typically owned by a packaged CDP.
• Infinity Campaigns: Continuous, goal-driven engagement loops replacing traditional one-time batch campaign runs. Agentic re-evaluation; closest KG analogue: latency-tier.intra-day or latency-tier.daily with an AI-driven orchestration loop.
• Lakehouse Federation: Query data across Databricks, Snowflake, and BigQuery without duplication. Extends Databricks relevance to Snowflake/BigQuery-primary organizations as a cross-warehouse querying layer (see vendor.snowflake, vendor.bigquery for alternatives).

Competitive implication. CustomerLake elevates vendor.databricks from composable CDP substrate to a direct competitor to packaged CDPs in Databricks-native enterprise accounts. For vendor.snowflake- or vendor.bigquery-primary organizations, vendor.hightouch and vendor.census remain the natural reverse-ETL activation layer — CustomerLake does not change that routing. Early customers (vendor-asserted): HP, Circle K, AB InBev, Getnet by Santander. Launch partners (21, vendor-asserted): Adobe, Braze, Bloomreach, Meta, and others.

Status qualifier: CustomerLake is in Private Preview as of June 2026; no GA date announced. All capability claims derive from CMSWire editorial coverage of Databricks marketing materials (source.cmswire-com.customer-experience-databricks-cdp-customerlake-2026). Apply the (vendor-asserted, Private Preview) qualifier to all CustomerLake claims until a second independent analyst source or official Databricks technical documentation confirms capabilities. Primary Databricks blog documentation now available (source.databricks-com.blog-introducing-customerlake-agentic-cdp-2026, Tier-1). Capability claims remain (vendor-asserted, Private Preview) pending independent analyst confirmation or GA announcement.

Where it fits. Organizations whose CDP ambitions extend into predictive scoring, recommendation models, or generative-AI personalization. Less suited to organizations whose marketing operations don't yet leverage ML.

Alternative-to. vendor.snowflake, vendor.bigquery (peer lakehouses/CDWs); vendor.adobe-experience-platform (packaged-CDP alternative).