Every quarter, enterprise marketing teams audit their campaign performance and arrive at the same familiar conclusions: open rates are declining, click-through rates are flat, and conversion attribution is murky at best. The instinct is to blame the email template, the subject line, the send cadence, or the marketing automation platform itself. But platforms like Oracle Eloqua, Adobe Marketo Engage, Salesforce Marketing Cloud, and HubSpot are, for the most part, doing exactly what they were designed to do. The dysfunction originates one layer down, in the fragmented, duplicated, inconsistently formatted customer data that feeds every campaign.
A recent analysis from Databricks sharpened this argument. Marketing teams, the company observed, blame their tools when campaigns fall flat. The real bottleneck, however, sits in customer data scattered across CRMs, email service providers, analytics warehouses, ad platforms, and a half-dozen other systems that never cleanly talk to each other. This observation is not new in itself. What makes it worth examining now is the degree to which the problem has compounded as enterprise campaign programs have grown more ambitious: more segments, more personalization tokens, more lifecycle stages, more channels. The gap between what campaign orchestration platforms can theoretically do and what they actually execute, given the data they receive, has never been wider.
1. Historical context
The separation between data infrastructure and campaign execution was, for many years, a manageable inconvenience. In the early 2010s, when marketing automation platforms first gained enterprise traction, most organizations ran relatively simple email programs. A handful of segments (prospects, customers, lapsed), a limited set of personalizations (first name, company, maybe industry), and a linear funnel model. The data requirements were modest. A CRM sync and a basic import process could keep things moving.
The shift happened gradually, then all at once. Between 2016 and 2020, account-based marketing emerged as a dominant strategy. Multi-touch campaign architectures replaced single-send blasts. Behavioral scoring models layered web activity, content engagement, and product usage data into lead qualification. Privacy regulations (GDPR in 2018, CCPA in 2020) added consent management as another data dimension. Suddenly, the modest CRM sync was insufficient.
Platforms responded by expanding their native integration capabilities. Salesforce Marketing Cloud added more connectors. HubSpot built Operations Hub. Eloqua and Marketo deepened their API ecosystems. But the integrations were, and largely remain, point-to-point. Each connection solves one data flow problem without addressing the systemic issue: there is no single, reconciled, continuously updated view of a contact or account that every campaign can draw from.
The customer data platform (CDP) category emerged in part to fill this gap. But as David Raab, founder of the CDP Institute, has noted, many CDP implementations end up creating yet another data silo rather than eliminating existing ones. The result, by 2025, is an enterprise marketing data architecture that looks less like a clean pipeline and more like a patchwork of partial truths.
"Most CDP implementations end up creating yet another silo rather than connecting the ones you already have."
2. Technical analysis
To understand why campaign performance suffers, you have to trace the data path from source to send.
Consider a typical enterprise running Marketo or Eloqua for email and campaign automation, Salesforce as the CRM, a separate analytics warehouse (Snowflake, Databricks, or BigQuery), a web analytics tool, and one or more ad platforms. A contact record might exist in all of these systems, but with different identifiers, different field values, and different update timestamps.
When a campaign team builds a segment for a product launch email, they are working with whichever version of the truth their marketing automation platform has. If the CRM sync runs nightly but the contact updated their preferences that morning, the segment is stale. If the analytics warehouse contains purchase history that has never been pushed into the MAP, the segmentation misses a behavioral signal. If the ad platform has suppression data that is not reflected in the email tool, the contact receives an irrelevant message.
The deduplication problem
Duplicate records remain one of the most persistent and damaging data quality issues in campaign operations. A 2023 Validity report found that 44% of organizations estimate their CRM data degrades at a rate of at least 25% per year. When duplicates flow into the MAP, they cause inflated segment counts, redundant sends, inconsistent personalization (one record might have a job title; the duplicate might not), and skewed engagement metrics. A contact who appears twice in a campaign audience and opens one email produces a 50% open rate at the contact level, a distortion that cascades into reporting and optimization decisions.
Data deduplication is not a one-time cleanup exercise. It requires continuous matching rules, merge logic, and source-of-truth hierarchies that most marketing teams do not maintain on their own.
The normalization gap
Field-level inconsistency is another silent campaign killer. When "United States," "US," "U.S.A.," and "America" all appear in the same country field, any segment built on geography will be incomplete. When job titles arrive as free text from web forms, building a reliable persona-based campaign becomes an exercise in guesswork. When industry codes differ between the CRM and the MAP, account-level targeting loses precision.
These are not exotic edge cases. They describe the default state of most enterprise marketing databases. Without systematic data normalization, every downstream campaign decision inherits the noise.
The integration architecture
The Databricks analysis points to a pattern that most enterprise marketing operations teams will recognize: data is not missing, it is marooned. The behavioral signals exist in the web analytics tool. The firmographic enrichment exists in the data warehouse. The consent records exist in the preference center. But no single system holds the complete picture, and the connections between systems are brittle, batch-oriented, or manually maintained.
This is why the conversation about campaign performance must shift from "which email platform should we use" to "how do our data systems feed the platform we already have." As our earlier analysis of broken MarTech stacks argued, the stack problem is almost always a strategy problem first.
3. Strategic implications
For enterprise marketing leaders, the data layer problem has three strategic consequences that extend well beyond email metrics.
Personalization hits a ceiling
Every major MAP vendor now offers advanced personalization: dynamic content blocks, conditional logic, AI-driven send-time optimization, predictive content recommendations. Adobe's Marketo Engage has GenAI-assisted content suggestions. Salesforce Marketing Cloud has Einstein engagement scoring. But these capabilities are only as good as the data that powers them. If your contact records lack reliable industry, role, or engagement history data, dynamic content blocks will render generic fallbacks. The platform will personalize beautifully for the 30% of records with complete data and deliver a degraded experience to the other 70%.
This ceiling is particularly painful for organizations investing in multi-touch campaigns and journey orchestration. A multi-stage nurture program that adapts based on behavioral signals cannot function when those signals are trapped in a system the MAP cannot access in real time.
Attribution becomes unreliable
Campaign reporting depends on the ability to connect a marketing touch to a revenue outcome. When the same person exists as two records in the MAP and a third in the CRM, attribution models break. Multi-touch attribution, which is already statistically fragile, becomes almost meaningless when the underlying identity graph is incomplete. The analytics maturity gap that many organizations experience is, in a significant number of cases, a data quality gap wearing a reporting hat.
Compliance risk compounds
GDPR, CCPA, and newer regulations like Canada's CASL require that organizations respect consent preferences at the contact level. When a contact's opt-out is recorded in the preference center but not propagated to every system that can trigger a send, the organization is exposed. Duplicate records make this worse: a contact opts out on one record, but the duplicate, which feeds a different campaign, remains active. This is not a hypothetical scenario. It is a recurring finding in privacy compliance audits.
Source: Validity State of CRM Data Health Report 2023
"The Martech industry is consolidating and expanding at the same time. There are now over 14,000 products, but the average enterprise uses only a fraction of the capabilities in the ones they already own."
4. Practical application
Resolving the data layer problem requires a structured approach. Here are concrete steps enterprise marketing operations teams can take.
Conduct a data flow audit
Before touching any technology, map every data source that feeds your campaign execution platform. For each source, document: what data it provides, how frequently it syncs, what format the data arrives in, and which fields are matched to records in the MAP. This exercise almost always reveals gaps, redundancies, and stale connections that the team did not know existed.
A platform maturity assessment can formalize this process, evaluating not just the platform's configuration but the quality and completeness of the data flowing into it.
Establish source-of-truth hierarchies
Not every system should have equal authority over every field. Define which system is the master for each category of data. The CRM might own account hierarchy and opportunity data. The preference center owns consent records. The MAP owns engagement data. The enrichment vendor owns firmographic fields. When conflicts arise, the hierarchy determines which value wins. Without this, merge operations and sync processes will overwrite good data with bad data on a recurring basis.
Implement continuous data hygiene
Batch cleanups are necessary but insufficient. Organizations that clean their database once a year (or once before a major campaign) are treating a chronic condition with acute-care interventions. Automated data enrichment and normalization processes, running on a regular cadence, prevent the re-accumulation of data debt.
Build a unified segment testing layer
Before any campaign launches, validate the segment against multiple data sources. If the MAP says there are 12,000 contacts in a target segment, cross-reference that count against the CRM and the data warehouse. Discrepancies indicate integration failures, duplicate inflation, or filtering logic that is too loose. This is a simple quality gate that most organizations skip.
Invest in real-time (or near-real-time) data pipes
Nightly batch syncs were adequate for weekly newsletter sends. They are not adequate for trigger-based campaigns, event-driven journeys, or behavioral personalization. If your campaign execution strategy depends on responding to what a contact did today, your data infrastructure must deliver that signal within minutes, not hours. Modern ETL solutions and reverse-ETL tools from vendors like Census, Hightouch, and Fivetran have made this achievable without rebuilding the entire stack.
5. Future scenarios
Over the next 18 to 24 months, three forces will accelerate the convergence of data infrastructure and campaign execution.
The composable CDP becomes the norm
The traditional CDP model, which requires organizations to copy all their customer data into a new centralized system, is losing favor. Composable CDPs (Hightouch, Census, Simon Data) sit on top of existing data warehouses and activate data directly into MAPs without creating another copy. This architecture reduces data drift, eliminates sync latency, and makes the warehouse the single source of truth for campaign targeting. For enterprise teams running Oracle Eloqua or Adobe Marketo, this means richer segmentation without the overhead of maintaining a separate CDP.
Gartner projected in 2024 that by 2026, 60% of large organizations would shift from buying standalone CDPs to composable architectures built on their existing data infrastructure. If that trajectory holds, the MAP becomes a pure execution layer, and the data warehouse becomes the brain.
AI-driven campaign optimization demands better data inputs
As AI features in MAPs become more capable (predictive send-time, next-best-action recommendations, automated content selection), the quality of the data feeding those models becomes a performance multiplier or a performance ceiling. An AI model trained on incomplete or duplicated engagement data will produce suboptimal recommendations. Organizations that have invested in data quality will see disproportionate returns from AI features. Those that have not will find their AI capabilities producing mediocre results and question the investment. As our analysis of predictive orchestration noted, the AI is only as intelligent as the data it consumes.
Privacy enforcement raises the stakes
Regulatory enforcement is tightening. The EU's enforcement of GDPR fines reached 4.2 billion euros cumulatively by the end of 2024, according to CMS Law's GDPR Enforcement Tracker. New regulations in the U.S. (state-level privacy laws in Texas, Montana, Oregon, and others that took effect in 2024) are expanding the compliance surface. For campaign teams, this means that every email send must be defensible at the contact level: provable consent, accurate preference data, proper suppression logic. Data fragmentation makes this nearly impossible. Organizations that do not solve their data unification problem will face compliance exposure as a direct consequence.
The role of marketing operations evolves
The MarTech Conference session referenced in the original news article frames MOps as shifting from a support function to a growth driver. That shift is real, but it depends on the data layer. An MOps team that spends 60% of its time on data troubleshooting, fixing broken syncs, deduplicating records manually, reconciling conflicting reports, cannot function as a strategic growth driver. Investing in the data infrastructure that eliminates this toil is the prerequisite for the MOps transformation that every CMO wants.
"Data quality is the silent killer of marketing automation ROI. You can have the best platform in the world, but if you're feeding it garbage, you're just automating garbage faster."
6. Takeaways
- Campaign underperformance in enterprise email programs is most often a data problem, not a creative or platform problem. The MAP executes what it is given; if the inputs are fragmented, stale, or duplicated, outputs will be poor.
- Duplicate records, inconsistent field values, and batch-only data syncs are the three most common data layer failures that degrade campaign results.
- Personalization capabilities in modern MAPs (Eloqua, Marketo, SFMC, HubSpot) are advancing faster than most organizations' data readiness to use them. The gap produces generic experiences despite enterprise-grade tooling.
- Source-of-truth hierarchies, continuous data hygiene, and real-time data integration are operational necessities, not nice-to-haves, for any organization running multi-touch or behavioral campaigns.
- Composable CDPs built on existing data warehouses will increasingly replace standalone CDPs, making the data warehouse the campaign targeting brain and the MAP the execution layer.
- AI-driven campaign optimization features will amplify data quality differences: organizations with clean, unified data will see multiplied returns; those without will see marginal or misleading results.
- Privacy compliance is becoming a forcing function for data unification. Fragmented consent data across disconnected systems creates regulatory exposure that grows with every send.
- The MOps function cannot evolve from tactical support to strategic growth driver until the data infrastructure stops requiring manual remediation. Fix the data layer first, and the team transformation follows.
