Why marketers are really cutting martech tools

Data architecture is now the main constraint in martech consolidation

Gartner expects 80% of net-new enterprise customer data platform (CDP) deployments to be embedded in or composable with data platforms by 2030. That forecast points to a structural change in marketing technology. Companies are reorganizing their stacks around where customer data lives, how applications access it, and how consistently the business governs it.

For years, consolidation focused on software inventories. Teams identified overlapping products, low usage, and unnecessary contracts. Those exercises can reduce spending. They leave a deeper problem unresolved when customer, campaign, and intent data remain distributed across systems with different schemas, permissions, and definitions.

Data fragmentation becomes more expensive as data volumes and use cases grow. Each application may maintain its own customer records. Integrations must then keep those records synchronized. A consent change, customer status update, or new transaction may reach different systems at different times. Teams spend more effort reconciling data, maintaining interfaces, and resolving conflicting definitions.

This effect is often called “data gravity.” Large and important datasets tend to become the center around which applications and computing workloads are organized because repeatedly moving and copying that data becomes costly and complex. TopicIntelligence identifies data gravity as an important force behind martech consolidation as companies introduce AI-driven decisioning.

AI increases the importance of this issue. An AI agent needs current, consistent, governed information to make reliable customer decisions. If customer identity, consent, purchase history, and campaign activity are spread across disconnected systems, the agent inherits those inconsistencies. Faster AI does not correct fragmented data.

Kabaleeswaran Sabapathi, Managing Enterprise Architect at Capgemini, describes the operational problem clearly: “Every new martech tool tends to pull its own copy of data instead of reading from a shared warehouse or operational data store.” He says this creates data silos, makes integration harder to manage, and contributes to low utilization. In his view, rising costs often start consolidation efforts, but governance eventually turns the discussion into one about data integrity.

The executive metric should therefore expand beyond licenses removed and dollars saved. Leaders should ask how many copies of customer data exist, which system holds the authoritative record, how quickly changes propagate, and whether governance policies remain consistent across applications. These questions expose structural complexity that a vendor count cannot show.

The direction is clear. Martech applications are becoming more dependent on a shared enterprise data layer. Companies that simplify access to governed customer data can reduce integration work while creating a cleaner foundation for AI, analytics, and consistent customer experiences.

Martech and CDP functions are converging around enterprise data platforms

Only 22% of marketers report high CDP utilization, while 41% of companies have implemented a CDP. Organizations also estimate that they use roughly 47% of the CDP capabilities they pay for. These figures expose a basic efficiency problem: many companies have bought more specialized functionality than their teams routinely use.

At the same time, traditional product boundaries are disappearing. Heinz Marketing identifies convergence across martech and revenue operations stacks. CRM platforms increasingly include marketing automation and sales engagement. Marketing automation products are expanding into CDP functions. Large suites continue to absorb capabilities that previously required separate point products.

This changes the economics of specialization. A dedicated product needs to provide enough additional business value to justify another integration, another data flow, another governance surface, and another vendor relationship. Feature quality alone becomes a weaker argument when an existing platform can already cover most of the workflow.

Customer data infrastructure is also moving closer to the enterprise warehouse or lakehouse. A data warehouse stores structured enterprise information for analytics and operations. A lakehouse combines warehouse-style management with the ability to handle broader forms of data. Both can provide a central environment from which marketing applications access governed customer information.

Databricks made this convergence explicit in June 2026 with the launch of CustomerLake, an agentic CDP built natively inside its lakehouse. The move placed a major enterprise data-platform provider directly into the CDP market. Gartner characterized the launch as evidence that customer data management is shifting closer to enterprise data platforms. Gartner also advised CMOs to evaluate the offering as a data infrastructure decision.

That principle applies more broadly to CDP purchasing. A CDP determines where customer records reside, how identities are resolved, which systems can access them, and how consent and permissions are enforced. It can also shape what data future AI systems can use. A renewal decision can therefore affect marketing, IT, security, analytics, data engineering, and AI teams for years.

Supriya Agarwal, Director of Marketing, Data and AI at SOSV, identifies the core operational constraint. “Inconsistent schemas, metadata and permissions, along with growing integration and access requirements, are pushing organizations toward consolidating around fewer, more integrated platforms.” She adds that this approach simplifies governance, improves interoperability, and creates a stronger AI foundation because AI performs best with structured, consistent, well-governed data.

C-suite leaders should consequently evaluate martech consolidation at the enterprise architecture level. Marketing still owns critical requirements around segmentation, activation, personalization, and campaign execution. Data and IT leaders need an equal role in decisions that determine data location, access controls, integration patterns, and AI readiness.

The strategic objective is a smaller number of authoritative data environments with flexible applications around them. That model gives companies more freedom to change marketing tools without repeatedly rebuilding the customer data foundation. It also directs investment toward the asset that every future marketing application and AI agent will need: trusted, governed customer data.

The CDP market is splitting into two models built around shared enterprise data

Gartner describes two emerging directions for customer data platforms: platformization and agentification. Both models increase the strategic importance of the underlying data layer. The main difference is where customer intelligence and execution take place.

Platformization puts the CDP at the center of a broader application suite. Vendors such as Adobe, Oracle, and Salesforce can connect customer data with marketing, analytics, personalization, advertising, sales, and service functions inside their respective ecosystems. This approach can reduce integration work for companies that already depend heavily on one vendor.

The tradeoff is architectural dependence. When customer data management and business applications are tightly coupled to the same suite, changing vendors can become harder. Executives should examine data portability, interoperability, governance controls, and the cost of replacing individual components. A broad feature set has limited strategic value if switching one application requires substantial changes to the customer data architecture.

Agentification takes a different approach. The CDP remains relatively thin, while autonomous software agents use a shared, governed data layer to make decisions and execute tasks. An agent could query customer history, preferences, consent status, and current behavior before deciding which interaction should occur. Applications then become execution endpoints for those decisions.

This model raises the standard for data quality. AI agents can operate quickly and at large scale. Their decisions therefore depend on reliable identity resolution, current consent information, consistent definitions, and traceable source data. An organization that has five conflicting versions of a customer record creates uncertainty before an agent makes its first decision.

Governance also changes when software agents participate in customer-facing decisions. Enterprises need to know which data an agent accessed, which rules it applied, what action it recommended or executed, and whether that decision can be traced to authoritative records. Human oversight remains important for consequential or poorly bounded decisions. Automation requires clear permissions and accountability.

For C-suite leaders, the platformization-versus-agentification decision should start with existing architecture. A business deeply invested in a major enterprise suite may gain operational simplicity from platformization. A company with a mature warehouse or lakehouse and strong data engineering capabilities may have more reason to consider a composable, agent-oriented model.

The durable asset in either strategy is governed customer data. Applications and AI capabilities will continue to change. A clean data layer gives the company more options to adopt those changes without repeatedly rebuilding its customer-data foundation.

A universal data layer improves operational efficiency and customer consistency

A universal data layer keeps customer information in one governed environment, typically a cloud data warehouse or lakehouse. Marketing applications access that environment when they need customer information. This design creates a common set of identities, permissions, attributes, and business definitions across channels.

The operational benefit is immediate. Marketing teams often spend substantial time reconciling reports because different platforms hold different versions of customer activity. One system may classify a person as an active customer while another uses an older definition or stale transaction history. A shared data layer allows applications to use the same underlying record and definitions.

Consent management shows why this matters. If a customer changes a tracking or communication preference, that update should become available to every relevant application quickly. Advertising, email, websites, analytics systems, and AI agents should all work from the current preference. Centralized governance reduces the number of independent records that must be synchronized and monitored.

This directly affects customer experience. Consider a shopper who withdraws tracking consent in a retailer’s application. An advertising system using stale data could continue retargeting that person. The company then creates an inconsistent experience and potentially increases privacy and compliance exposure. A governed data layer makes the latest preference available across connected systems.

The same principle applies to personalization. Product recommendations, loyalty status, campaign eligibility, and service interactions improve when every channel can access the same customer state. A recent purchase can inform the next email. A loyalty-tier change can become available to the website and customer service environment. Marketing execution becomes more coherent because applications make decisions from common information.

Centralization still requires disciplined governance. A universal data layer does not automatically produce accurate data. The business needs clear customer definitions, identity-resolution rules, access controls, metadata, quality checks, and ownership. It also needs explicit service expectations for how quickly important events and consent changes become available to downstream systems.

Executives should therefore judge a unified-data strategy by operational outcomes. Useful questions include how many customer-record copies remain, how quickly changes propagate, whether applications use consistent business definitions, and whether decisions can be traced to their underlying data. These measures reveal whether consolidation has produced a genuinely shared operating environment.

The strategic benefit extends into AI. An AI agent performs better when it can query one governed customer record containing current permissions and relevant context. A universal data layer creates that foundation while reducing the reconciliation and synchronization work that has historically consumed marketing and IT resources.

Warehouse-native composable architectures provide the strongest foundation for governance and AI

Three common architecture models create very different outcomes for data duplication, governance, and AI readiness: point-to-point integrations, packaged CDPs, and warehouse-native composable systems. The key variable is how many copies of customer data each architecture creates and where governance is enforced.

Point-to-point integration connects applications directly. This approach can be effective with a small number of systems. Complexity rises quickly as the stack grows. Each new application can require several additional integrations, along with separate mappings for customer identities, schemas, permissions, and events. The resulting environment has high data duplication, inconsistent governance, and fragmented context for AI.

Packaged CDPs address part of this problem by centralizing customer information inside a dedicated platform. They provide capabilities such as identity resolution, segmentation, audience creation, and activation. Governance becomes more centralized within the CDP. However, customer information may already reside in an enterprise warehouse, so the CDP can create another managed copy that needs to remain synchronized with upstream systems.

Warehouse-native composable architectures change this model. Customer information stays in the warehouse or lakehouse, while marketing applications query and activate that data from its existing location. Zero-copy access can reduce or eliminate additional physical copies for supported workflows. Governance policies can then be enforced closer to the authoritative data.

This architecture receives the strongest AI-readiness assessment in the comparison presented for 2026 martech stacks. Point-to-point integration is rated low for AI readiness because context remains fragmented across systems. Packaged CDPs receive a moderate rating that depends partly on connector coverage. Warehouse-native composable systems receive a high rating because AI agents can query a governed customer record directly.

The difference matters because AI decisioning requires context. An agent deciding whether to send an offer may need purchase history, loyalty status, recent behavior, product availability, communication preferences, and consent information. Pulling those attributes from several independently synchronized systems increases the chance of stale or contradictory inputs.

A composable architecture also changes the responsibility of the activation layer. The activation product can query customer data, apply segmentation or decisioning logic, and send instructions to email, advertising, web, or other channels. The enterprise data environment remains responsible for the underlying record.

Zero-copy should still be evaluated as a technical capability rather than accepted as a vendor label. Executives should determine which workloads actually operate without copying data, where temporary data is stored, how permissions are inherited, and what happens when applications require capabilities unavailable in the warehouse. These details determine whether the architecture delivers the expected governance benefits.

For companies with a mature warehouse or lakehouse, a warehouse-native model can create a strong base for AI and marketing activation. The business gains the most when data engineering, governance, security, and marketing teams share clear ownership of the architecture.

Every additional customer-data copy increases governance risk and AI uncertainty

Data duplication creates a specific management problem: every copy can develop a different version of the customer. The risk grows when records move among CDPs, CRMs, advertising platforms, analytics tools, marketing automation systems, and data warehouses on different update schedules.

Consent is one of the clearest examples. A customer may withdraw permission for a particular use of their data. The authoritative system records the change immediately, while a downstream application may continue operating from an older copy. The organization then has to control both the policy and the speed at which that policy propagates through its technology stack.

Business definitions can drift in the same way. Teams may use different rules for terms such as “active customer,” “high-value customer,” or “qualified lead.” When those definitions are encoded independently across applications, two systems can reach different conclusions about the same person even when both systems function as designed.

AI increases the operational consequences. Automated systems can make large numbers of decisions quickly. If an AI agent receives stale consent status, incomplete transaction history, or conflicting customer attributes, those data problems become inputs to its decisions. Better models cannot resolve an authoritative-data problem on their own.

The relationship between duplication and AI readiness is visible across the three architecture models used for the 2026 stack comparison. Point-to-point architectures have high duplication, inconsistent governance, and low AI readiness. Packaged CDPs have moderate duplication, centralized CDP governance, and moderate AI readiness. Warehouse-native composable layers have low-to-no duplication through zero-copy access, source-level governance, and high AI readiness.

This makes data lineage increasingly important. Leaders need to know where a customer attribute originated, when it was updated, which transformations were applied, and which applications or agents used it. Traceability becomes especially important when AI systems influence customer-facing decisions.

Reducing duplication also simplifies operational ownership. Fewer customer-record copies mean fewer synchronization processes, permission models, schema mappings, and failure points to maintain. This can reduce reconciliation work and make accountability clearer when data quality problems occur.

The executive objective is therefore measurable. Track how many persistent copies of critical customer information exist, how quickly important changes reach downstream applications, and how often systems disagree on core attributes. Add consent propagation time, data freshness, and decision traceability to the governance scorecard.

AI readiness begins with reliable inputs. A company that controls customer identity, permissions, definitions, and data freshness creates a stronger basis for automation. Reducing unnecessary duplication is one of the most direct architectural steps toward that goal.

Treat every CDP renewal as an enterprise infrastructure decision

A CDP renewal determines more than which marketing features remain available. It can affect where customer data resides, how applications access it, how consent is enforced, and what information AI systems can use. These decisions reach across marketing, data, IT, security, privacy, analytics, and AI.

Gartner’s forecast shows the direction of travel. By 2030, Gartner expects 80% of net-new enterprise CDP deployments to be embedded in or composable with data platforms. Databricks reinforced this shift with the June 2026 launch of CustomerLake, an agentic CDP built natively inside its lakehouse. Gartner characterized that move as evidence that customer-data management is moving closer to enterprise data platforms and advised CMOs to approach CustomerLake as a data infrastructure decision.

That changes how companies should manage the buying process. Marketing teams should define business requirements such as segmentation, personalization, journey execution, and channel activation. Data and IT teams should assess architecture, interoperability, security, governance, and operational ownership. Privacy specialists should verify how consent and customer preferences move through connected systems.

Zero-copy access deserves specific attention. Vendors increasingly promote architectures that query data directly from a warehouse or lakehouse. Buyers need to establish whether customer records stay in their governed environment during real workloads, which functions create additional copies, and how access controls work across applications.

Consent propagation is another practical test. A platform should be able to show what happens after a customer changes a preference. Executives need to know which systems receive that update, how quickly they receive it, and how failures are detected. This makes governance measurable through operational behavior.

AI adds another requirement: traceability. If an AI agent recommends or executes a customer-facing action, the organization needs visibility into the information used for that decision. Decision logic, permissions, and relevant source data should be traceable. This is particularly important as companies give agents greater autonomy over personalization and activation.

A limited pilot can test these claims before a large migration. An abandoned-cart workflow or loyalty-tier update provides a contained use case. Run it through a warehouse-native activation path for one quarter. Compare data freshness, synchronization effort, maintenance workload, consent handling, and execution reliability with the existing process.

This approach turns architecture claims into observable results. It also gives executives evidence about operational impact before making a larger commitment.

The renewal decision should therefore start with a simple question: what long-term data architecture will this contract create? Contract price and feature coverage remain relevant. Data ownership, governance, interoperability, and AI readiness determine the wider enterprise impact.

Customer-record duplication is a better measure of martech complexity

Vendor count is easy to measure. The number of places where customer data must live is more useful for understanding architectural complexity.

A company can reduce its martech vendor count and still maintain several versions of the same customer record. Those records may sit in a warehouse, CRM, packaged CDP, email platform, advertising system, and analytics environment. Every persistent copy introduces synchronization, permissions, schema management, monitoring, and governance work.

The executive metric should therefore focus on customer-record distribution. Ask how many persistent copies exist before a marketing application or AI agent can make an accurate decision. Then measure how quickly important changes propagate and how frequently systems disagree on critical attributes.

Consent makes the business impact clear. A customer may change a tracking or communication preference in one channel. That preference needs to reach every relevant execution environment before the next decision occurs. Multiple independently maintained customer records increase the coordination required to achieve that result.

Customer definitions create a similar issue. If one platform considers a customer active after a purchase within 30 days and another uses 90 days, the organization has two operational definitions driving decisions. A governed data layer can establish shared definitions and make them available across applications.

AI makes this measurement more important. An autonomous agent may need identity, consent, purchases, behavior, loyalty status, and campaign history within seconds. Accurate decisions depend on current and consistent context. Fragmented records increase the amount of reconciliation required before the agent can act reliably.

The 2026 architecture comparison supports this relationship. Point-to-point integration is associated with high duplication, inconsistent governance, and low AI readiness. Packaged CDPs have moderate duplication and AI readiness, with governance centralized within the CDP. Warehouse-native composable architectures have low-to-no duplication through zero-copy access, source-level governance, and high AI readiness.

Executives can turn this into a compact set of operating measures: persistent customer-record copies, consent propagation time, data freshness, conflicting attribute rates, synchronization failures, and the percentage of decisions traceable to governed data. These measures connect technical architecture to customer experience, risk, and operational efficiency.

The broader objective is controlled data movement. Some copies will remain necessary for performance, resilience, regulatory requirements, or the technical needs of individual applications. The goal is to make each persistent copy intentional, governed, and economically justified.

That provides a stronger definition of successful consolidation. A modern martech stack should give applications and AI systems reliable access to governed customer information with minimal duplication. Fewer synchronization paths reduce operational complexity and create a cleaner foundation for consistent customer decisions.

Final thoughts

Martech consolidation should leave the business with a simpler data architecture. Cutting five vendors while keeping five conflicting customer records solves little. The stronger outcome is fewer unnecessary data copies, consistent governance, and reliable access to current customer information.

For executives, the next CDP renewal deserves the same scrutiny as any major data infrastructure decision. Ask where the authoritative customer record lives. Test how quickly consent changes reach every channel. Determine which workloads require copied data. Require AI decisions to be traceable to governed information.

AI raises the stakes. Agents can make customer decisions faster and at greater scale, which makes data quality, permissions, and freshness more consequential. A fragmented data foundation will carry its inconsistencies into automated decisions.

The practical goal is clear. Build around governed customer data and keep the application layer flexible. That gives marketing room to change tools as requirements evolve while preserving the data foundation that customer experience, analytics, and AI increasingly depend on.