CDP Consolidation: Audit Data Ownership Before Renewal

The CDP market is not consolidating around one winning vendor. It is consolidating around the place where customer data is governed.

Before renewal, note the direction of travel from Databricks, Zeotap, and Celebrus: the CDP is becoming a control layer over existing data, not another database marketing must populate and reconcile.

The CDP is moving to the data boundary

Scott Brinker’s 2026 landscape makes the pressure visible. The market now contains 15,505 products, up just 0.79% year over year (chiefmartec). That flat headline hides churn: vendors are being bundled, replaced, or pushed into narrower jobs while buyers ask harder questions about duplicated data, governance, and integration cost.

The practical response is not “buy fewer logos.” It is to decide which system owns each customer-data operation: collection, identity resolution, segmentation, decisioning, activation, and measurement. A CDP that cannot answer that clearly is a second pipeline with a nicer interface.

Databricks’ CustomerLake is the boldest version of this shift. Announced in private preview on June 16, the product places customer data, identity resolution, audience creation, campaign automation, and activation inside the Databricks lakehouse, governed by Unity Catalog (Databricks announcement). The key change is the proposed removal of a separate customer-data copy.

Zeotap is taking a similar architectural route from the Snowflake side. Its Composable CDP is available as a Snowflake Native App, with identity resolution, unified profiles, segmentation, journey orchestration, and activation running inside the customer’s Snowflake account (Zeotap product release). That changes the procurement conversation from “which CDP has the most connectors?” to “which capabilities should run inside our governed environment?”

What the releases actually change

These products do not eliminate integration work. They move the integration boundary.

Celebrus’ latest CDP update makes that explicit. The Celebrus API Connector adds a no-code interface for connecting external applications and generating the code for a functional API. The same release includes Visibility Detection 2.0, designed to measure whether targeted content was actually visible and engaged with—not merely sent (Celebrus release).

That is a useful counterweight to warehouse-native enthusiasm. A governed data layer is valuable only if signals can leave it, arrive in downstream systems, and return with enough context to support measurement. Celebrus also cites a Forrester survey in which 61% of organizations, according to Forrester (survey cited by Celebrus), found capturing and making sense of digital customer data difficult; 55% of organizations, according to Forrester (same survey), said understanding customers across touchpoints was difficult. The implication is blunt: consolidation without observable data movement simply hides the failure.

This is why the recent reverse-ETL discussion matters. A warehouse-native activation model can reduce duplicate storage, but it does not automatically solve identity quality, destination contracts, consent, or feedback loops. The operator question is still whether a conversion, audience membership, or content-exposure signal can be traced from source to action and back to revenue. Reverse ETL is quietly replacing the MQL handoff covers that operating shift in more detail. For the identity and delivery side, Server-Side CAPI Is Only Half the Fix shows why a successful handoff still needs match quality and observability.

Run a renewal audit before you compare features

Before accepting a CDP renewal proposal, create a capability-to-data-flow map. For every feature, record the source table or event stream, the transformation or identity rule, the destination, the API or connector used, the owner, and the proof that the downstream action occurred.

Then ask five uncomfortable questions:

Are we paying a vendor to store a copy of data our warehouse already governs?
Which identity rules are portable if we change vendors?
What happens when a destination API changes or silently drops a field?
Can marketing operations inspect and replay failed syncs without engineering?
Does the platform measure exposure and outcome, or only delivery?

Use those answers to split the stack into three buckets: capabilities to keep in the CDP, capabilities to move into the warehouse, and capabilities that should be removed. This is a better renewal strategy than comparing connector counts or accepting a bundled “platform” discount.

The Monday test

Ask your team to trace one high-value audience from its source record to a live activation and back to a measured outcome. If that walkthrough requires a spreadsheet, a ticket, or an undocumented custom script, the stack is not consolidated—it is merely distributed behind fewer brand names.

The winning vendor will be the one that makes ownership, movement, and failure visible. Choose on that basis, and renewal becomes a data-governance decision with marketing consequences—not another logo negotiation.