Your data warehouse is a critical investment. So is your CDP. The question isn’t which one wins—it’s how they work together to drive outcomes.
In MarTech and CX, change is constant. The past 18 months have been no exception. At Oracle, we regularly speak with some of the world’s largest organizations about the transformations happening across AI and data platforms.
Whether driven by new AI agent strategies, cost optimization efforts, or long-standing questions like data ownership, one theme remains consistent: trusted, unified, and up-to-date first-party data is essential for business outcomes. Two technologies have emerged as foundational to this strategy: the data lake/warehouse and the customer data platform (CDP).
So when I hear, “Shouldn’t I just use my data warehouse as my CDP?” It’s a fair question.
At Oracle, we deeply understand the strategic value of the data warehouse in customer data management. In our own environment, we integrate Oracle Unity Data Platform with the Oracle Autonomous AI Lakehouse to unify, manage, and activate enterprise-wide first-party data.
At the same time, terms like “composable CDP” and “zero-copy” are everywhere—and often misunderstood. While these approaches can simplify data architecture, they fall short when treated as a complete CDP strategy rather than one tool within a broader approach.
When organizations make that mistake, they risk optimizing for data location instead of business outcomes—the very reason a CDP exists.
So let’s explore where zero-copy works, where it doesn’t, and what an outcome-driven strategy really looks like.
What Zero-Copy Does Well
Zero-copy (or composable) architectures offer real advantages in the right scenarios:
- Governance and compliance. Keeping data in the warehouse ensures existing controls (access, encryption, and residency) apply automatically. This is especially valuable in regulated industries.
- Reduced data duplication. Eliminating redundant datasets lowers storage costs and reduces version conflicts. No more debating which system holds the “correct” customer record.
- Leverages existing investments. Organizations with mature data warehouses can activate curated data without replicating it into another platform.
These are meaningful benefits and shouldn’t be dismissed.
Where the Cracks Appear
Many high-value CDP use cases don’t behave like analytical queries. They require low latency, real-time responsiveness, and persistent state.
- Agentic AI and automation. Modern marketing and sales workflows increasingly rely on embedded AI agents. These systems need sub-second access to precomputed profiles and insights. Querying a warehouse for every decision introduces too much friction.
- Real-time triggers and campaigns. Moments matter. Whether it’s a pricing page visit or a chatbot interaction, response windows are measured in seconds. Real-time personalization and triggered campaigns require locally available profiles, not live warehouse queries.
- Identity resolution at scale. Stitching together identities across devices, channels, and systems is compute-intensive. Doing this on-demand via warehouse queries increases both latency and cost. Pre-resolved identity graphs are far more efficient.
- B2B buying group modeling. In B2B, the “customer” is a group, not an individual. Modeling relationships across accounts, contacts, and roles requires persistent data structures closer to activation—not ad hoc queries against raw tables.
The Hidden Cost Nobody Budgeted For
There’s a practical consideration that often surfaces only after deployment: warehouse compute costs.
Every query a zero-copy CDP sends back to the data warehouse consumes compute resources. Audience builds, segment refreshes, profile lookups, and campaign triggers all spin up warehouse capacity. Industry research suggests that moving from daily audience refreshes to hourly can increase compute costs by 25x, and pushing toward near-real-time refreshes can drive costs up by 50x or more.
Critically, these costs don’t show up on the CDP vendor’s invoice. They appear on the data warehouse bill, often catching IT and finance teams off guard. The more use cases an organization activates through a zero-copy model, the higher the warehouse compute load becomes. What started as an architecture designed to reduce costs can end up generating significant new ones in a different line item.
The Right Integration Method for the Right Use Case
The most effective CDP architectures don’t commit to a single integration pattern. They match the integration method to the requirements of each use case.
- Zero-copy / federated query works best for governance-sensitive reference data, enrichment attributes that change infrequently, analytical workloads, and scenarios where the warehouse already serves as the operational system of record.
- Batch ingestion is the right choice for large-volume data loads (historical transaction data, CRM syncs, third-party enrichment files) where freshness requirements are measured in hours rather than seconds.
- Real-time streaming is essential for behavioral event data, website interactions, product usage signals, and any trigger-based activation that demands sub-second response.
- Precomputed profile persistence is necessary for identity resolution, buying group models, AI-driven scoring, and any use case where the activation layer needs instant access to a resolved, enriched customer record without re-querying the warehouse.
The question to ask isn’t “Can we avoid copying data?”
It’s “What does each use case need to perform at its best, and what integration pattern delivers that?”
Why This Matters for B2B
B2B adds complexity. Unlike B2C, where segmentation may rely on a single profile, B2B requires:
- Account hierarchies
- Buying group mapping
- Multi-contact engagement tracking
- Sales + marketing alignment
These require purpose-built data models and native entity management, not just identity stitching. When “composable” approaches push this responsibility onto internal teams, the hidden cost is engineering effort. What looks simple on paper can become expensive in practice.
A Practical Framework for Evaluation
For organizations evaluating CDP architecture, we recommend this approach:
Start with use cases, then architecture. Define the five to ten activation scenarios that matter most (real-time triggers, buying group engagement, predictive lead scoring, campaign orchestration, sales alerts, etc) and document the data freshness, latency, and compute requirements for each.
Map integration methods to requirements. For each use case, determine whether zero-copy, batch, streaming, or precomputed persistence is the best fit. Most enterprises will need a mix.
Quantify the full cost picture. Model not just software licensing but warehouse compute costs under load, internal engineering hours for build-and-maintain scenarios, and the opportunity cost of delayed activation when architecture limits real-time response.
Evaluate native B2B capabilities. If you sell to businesses, ask whether the platform delivers buying group modeling, account hierarchy management, and commercial object mastering natively — or whether your team will need to build those from scratch.
Plan a workshop, not a proof of concept. Before committing to an architecture, bring data engineering, marketing operations, and commercial leadership together to align on which data needs to be locally persisted for activation and which can be referenced via federated query. That alignment prevents costly rearchitecting down the road.
Bottom Line
Zero-copy is a valuable tool. But it’s not a strategy. What matters is whether your CDP drives pipeline, conversion, and retention.
That requires the right data, in the right place, at the right time and sometimes, that means moving data.
The best architectures:
- Respect the data warehouse / data lake as foundational
- Recognize the unique demands of activation workloads
- Balance flexibility with performance
Your architecture should serve outcomes. Not the other way around.
Learn more:
Oracle Fusion Unity Data Platform
Oracle named a Leader in the 2026 Gartner® Magic Quadrant™ for Customer Data Platforms
Webinar On-Demand: Building Exceptional AI-Driven Customer Experiences
Explore the Oracle AI Playbooks to help you on your AI journey