CDPs Failed. Now What? The Next Wave of Customer Data Strategy

The obituary for the customer data platform is too tidy for what is actually happening in the market. The category is still alive, and by the middle of last year the Customer Data Platform Institute counted 208 vendors and estimated industry revenue at about $2.6 billion. But the same industry update also showed the center of gravity moving away from the old stand-alone model and toward larger platforms, embedded capabilities, and acquisition-driven consolidation. In other words, the need for unified customer data did not disappear. The product story changed around it.

That shift is happening against a backdrop of extraordinary martech sprawl. Scott Brinker reported that the marketing technology landscape grew again in 2025 to 15,384 solutions, even as the previous generation of vendors consolidated. For marketing and data leaders, that has changed the core question. It is no longer whether customer data needs to be unified. It is where the trusted profile should live, how little data should be copied, how fast signals need to move, and which layer should turn data into audiences, journeys, decisions, and now AI actions.

The promise outlived the product

Under the original industry definition, a CDP was packaged software that created a persistent, unified customer database that other systems could use. That definition mattered because it described not just a business goal but an architecture. In the classic model, the platform copied data from source systems into its own store, resolved identities there, and then exposed audiences or profiles to downstream tools. For an earlier era of CRM, web analytics, email, and adtech fragmentation, that was a practical breakthrough.

What broke was not the ambition but the assumption that one packaged application could continue to sit in the middle of everything as data volumes, privacy pressure, channel complexity, and AI expectations all rose at once. Salesforce now promotes zero-copy access precisely because many enterprises do not want yet another duplicated data store. Adobe is talking about federated datasets, flexible access, and minimizing duplication. Databricks is pitching a unified data and AI foundation for marketing. The market is moving from one more database to a coordinated data foundation with multiple services running on top of it.

Recent executive language makes the change hard to miss. At Twilio, Khozema Shipchandler said businesses now need communications, contextual data, and AI as shared infrastructure for customer engagement. At Salesforce, Rahul Auradkar has described Data Cloud as an activation layer for enterprise AI. At Adobe, Anjul Bhambhri called agentic AI “a major leap forward” and tied it directly to reliable, real-time experience insights. None of those are the words of vendors defending the old monolith. They are the words of vendors rebuilding the middle of the stack around data access, governance, and activation.

Where so many projects broke

Many CDP projects underdelivered because they were sold as software deployments but behaved like enterprise operating-model change. Teams attempted to ingest every source, clean every identifier, and satisfy every stakeholder before proving a few hard business use cases. On a recent Twilio podcast, Glenn Vanderlinden, co-founder of Human37, said companies should “start with use cases” and warned that otherwise they end up with data that has no use case at all. That is a concise diagnosis of how many programs drifted into abstraction.

The technical failure modes were just as familiar. Identity stitching sounds elegant in a sales presentation, but in practice it means agreeing on schemas, match rules, key hierarchies, consent logic, and event standards across products, regions, and legacy systems. Adobe’s own implementation guidance shows how consequential those choices are. Profile enablement changes the behavior of the data model, and schema enablement for profile processing cannot simply be reversed. That is not a plug-in. It is foundational data engineering with real downstream consequences.

Then there was the duplication tax. Every copied profile created another place to manage personally identifiable information, another pipeline to watch, another mismatch between warehouse truth and activation truth, and another latency window in which the customer changed before the profile did. Salesforce’s zero-copy documentation now explicitly argues that teams can use identity resolution and segmentation without physically copying data into a separate system, while Adobe’s own summit guidance has begun framing federated datasets and low-duplication workflows as essential to privacy-centric activation.

The economics also caught up with the category. In early 2024, Twilio said it had conducted an operational review of Segment after recent underperformance, explored alternatives including a sale, and ultimately kept the business while focusing on faster onboarding, shorter time to value, and stronger data warehouse interoperability. That episode was broader than one company’s quarterly discipline. It was a signal that even one of the best-known names in the category had to confront how difficult and expensive the stand-alone middle layer had become.

What is replacing the monolith

The next wave is not anti-CDP. It is post-monolithic. In practice, companies are pulling apart the old package into layers: data collection lands first in the warehouse or lakehouse, identity resolution runs either there or through a connected graph, consent and policy stay closer to the source, and activation happens through audience tools, journey systems, reverse ETL, APIs, and increasingly AI assistants or agents. The effect is less a single product than a first-party data platform with modular services around it.

That architectural shift is already visible in how major vendors describe their marketing products. Databricks launched Data Intelligence for Marketing as a unified data and AI foundation, and Rick Schultz said it gives marketers “real-time, conversational analytics.” Adobe is using Experience Platform as the foundation for audience, journey, experimentation, and insights agents. Twilio is folding CDP capabilities into a broader customer engagement platform. The category is being reinterpreted as an activation layer on top of shared, governed data, not as a sealed box that must own every profile itself.

The distinction looks like this, based on how the legacy category was defined and how the current market is now being built.

Not every company needs the most composable version of this model. Many still prefer a packaged suite when data engineering capacity is thin or use cases are narrow. But the momentum is unmistakable. The fastest-growing patterns are the ones that use the data foundation a company already has, reduce copies, and keep activation flexible. Even Hightouch’s own argument for composable CDPs now includes an important caveat: most business use cases do not require true real-time, and some actually perform better when batch processing improves completeness, identity match rates, and auditability.

How brands are rebuilding the stack

The strongest real-world examples now come from companies treating customer data as a platform, not as a marketing island. On Snowflake, Armen Rostamian, vice president of marketing intelligence and analytics at BODi, said the company could finally see “the customer’s full interaction with the brand at scale.” Ralph Munsen, chief information officer at Warner Music Group, said the platform helps the company understand fan behavior and informs investment decisions around artists and content. These are not stories about buying a profile database for its own sake. They are stories about turning governed data into operating insight.

On the lakehouse side, Databricks says global brands are already using its data and AI foundation as part of broader martech stacks. Kumar Ram, global head of marketing data sciences at HP Inc., said dormant first-party data became useful once the company paired Databricks with ActionIQ. Tino Tomasone of PetSmart said the retailer now syncs millions of records into Salesforce Marketing Cloud every day, powers thousands of audiences, and sends more than four billion emails a year across a loyalty program with more than sixty-five million members. Manish Agarwal of Skechers said ActionIQ on Databricks has given marketers stronger data and analytics to improve lifetime value.

Salesforce is pushing the same direction from the application side. Its platform now presents zero-copy connectivity as a way to use identity management, segmentation, and analytics without physical duplication, and the company says enhanced bi-directional zero-copy capabilities are driving near real-time activation. In late 2025, Andy McCann of Yamaha Motor Corporation said the stack around Data 360 and Informatica would help build a “unified, trusted data ecosystem” across the business. Scott Strickland of Wyndham Hotels & Resorts said the same approach was reducing inconsistent definitions and manual effort.

Across those examples, the pattern is consistent. CDP functions have not disappeared. They have been redistributed across the enterprise data foundation, identity logic, privacy layer, and activation tooling. The profile now matters most when it is trusted, shareable, and actionable across teams, not when it sits in one more application.

Why vendors are racing to reposition

The vendor landscape is moving just as quickly as the architecture. The CDP Institute counted six CDP acquisitions in the first half of 2025 alone, matching the total from the prior two years. The biggest was Salesforce’s roughly eight billion dollar deal for Informatica. Marc Benioff called the combination an “ultimate AI-data platform,” a phrase that says as much about the future of CDPs as it does about one acquisition.

Other deals underline the same thesis. Uniphore bought ActionIQ and Infoworks to build what it calls a Zero Data AI Cloud. Chief executive Umesh Sachdev framed it as a way to remove data bottlenecks while preserving the customer’s existing landscape, and Tasso Argyros said bluntly that “the future of CDPs is AI and CX Agents.” Rokt’s merger with mParticle carried a similar message, with Michael Katz saying brands want to activate data in real time while retaining ownership and control.

Overlay that with the broader software market and the direction becomes even clearer. The CDP Institute says firms for which CDP is not the primary business now account for over a third of companies in the category and nearly half of estimated employment. Composable vendors continue to grow faster than conventional CDP vendors, even if their employment share remains relatively small. What looks, from a distance, like CDP failure is more accurately a transfer of power from stand-alone systems to broader data, activation, and AI platforms.

What marketing and technology leaders should do now

The practical response for CMOs and CTOs is not to begin with a rip-and-replace program. It is to get brutally specific about use cases, latency, and ownership. Start with suppression, loyalty, audience activation, churn, service personalization, and measurement. Ask where the most trusted first-party data already lives. Decide which identities truly require real-time stitching and which can be resolved in deliberate batches. Treat any new customer data strategy as a value program first and a platform choice second.

From there, separate the layers. Keep collection and core modeling close to the warehouse or lakehouse. Make identity resolution explicit, with documented match rules and confidence logic. Put consent, data minimization, encryption, masking, and access policy near the source. Use activation tools as services against that foundation, not as new systems of record. Where an existing CDP is deeply embedded, keep it alive for current channel orchestration while progressively moving profile logic, audience definitions, and measurement toward the shared data layer or a federated access model.

Governance, finally, has to be joint. The future-proof model is a co-led operating group in which marketing operations owns use-case prioritization, the data team owns productized datasets and quality service levels, privacy and legal own policy, and channel owners remain accountable for outcomes. Adobe’s summit programming now explicitly stresses co-led IT and marketing centers of excellence, while Salesforce is building policy-based governance, AI tagging, encryption, masking, and cross-source enforcement directly into Data 360. The era of customer data as a single marketing purchase is over. The next era belongs to companies that treat it as shared enterprise infrastructure with activation at the edge.

The next wave of customer data strategy, then, is not a new acronym. It is a more disciplined architecture. Profiles are becoming less important as objects to be stored in one vendor’s box and more important as governed context that can move safely and quickly across marketing, commerce, service, and AI. The brands that adapt fastest will not be the ones that buy the loudest replacement for the old CDP pitch. They will be the ones that make first-party data trusted, low-duplication, privacy-aware, and close enough to activation that marketers can act without waiting weeks for a new audience, a new model, or a new answer.