Data has never received more executive attention. Organizations are actively pouring money into data and AI, boards are demanding answers, and CEOs expect ROI. Yet Chief Data Officer (CDO) tenures are unusually low.

According to a widely cited 2025 research report by the Data & AI Leadership Exchange, the majority of CDOs remain in the role for fewer than three years.

By comparison, a study by executive search and consulting firm Spencer Stuart found that C-suite leaders typically stay in their roles for five years or more. While the study did not include CDOs, it identified Chief Communications Officers as the shortest-tenured leaders, with an average term of 4.7 years.

If data is becoming more important and the role is transformational, why are the leaders responsible for it leaving so quickly?

The question matters because most data transformations take far longer than a few years to deliver value. Building trusted data, changing behaviors, reducing risk, and establishing effective AI governance are not short-term projects. They are long-term organizational capabilities.

To understand what makes a long-lasting CDO, we spoke to Justin Heller, former and inaugural CDO of Synchrony Financial. Heller spent more than a decade in the role at Synchrony, helping build the company’s enterprise data management function.

Across a series of conversations, six themes emerged. They were centered on leadership, process, governance, adoption, business alignment, and the ability to create sustainable change.

The lessons that follow offer a practical roadmap for current CDOs, aspiring data leaders, and executive teams seeking to understand what it takes to build a data leadership career and a data management function that lasts.

How to position data management as an operating capability

One of the first things Heller challenges is the idea that data management has a finish line. When organizations hire a CDO, there is often an assumption that the leader will come in, fix data quality issues, establish governance, implement controls, and eventually complete the transformation.

Heller believes that expectation creates problems almost immediately. Early in his tenure at Synchrony, he recalls being asked during a board discussion when the work would finally be done.

His answer was simple: Data management is not a project. It’s an operating capability.

That distinction influenced how Heller approached everything from governance and risk management to organizational design. Rather than building a collection of isolated initiatives, he focused on creating a function that could become part of the company’s operating model.

The temptation in that situation is to attack visible problems first. For many organizations, that means data quality. Heller took a different view: “If you accept the premise that all data is broken, you’ve already lost.”

In his experience, many so-called data quality issues are actually process issues. The implication is significant. If the problem sits inside the process, fixing the data will not solve it. However, improving the process might.

That mindset shaped how Heller prioritized investments and communicated value across the organization. Instead of positioning data management as a cleanup exercise, he positioned it as a way to improve how the business operated.

As Heller puts it: “I had the unenviable task of running something that never ends.”

Read more: CDO Tenure – How to Succeed as a Long-Term Chief Data Officer

Why process beats technology

Most data conversations eventually drift toward technology. Someone is evaluating a new governance platform. Someone else is implementing a catalog. Another team is exploring AI-powered metadata management.

Heller doesn’t dismiss any of those investments, he just doesn’t start there.

When establishing Synchrony’s data management capabilities, one of his first priorities was creating clarity around ownership. Data was organized into domains such as customer, finance, and product. Critical data elements were identified. Subject matter experts were assigned responsibility.

The objective wasn’t to create bureaucracy. It was to answer a simple question: who understands this data best?

Heller’s approach was to align ownership to expertise. If there was disagreement about where a data element belonged, the discussion wasn’t resolved through mandates. It was resolved through dialogue and consensus.

Another principle that shaped the program was risk-based decision-making. Heller prioritises understanding the risk of not doing something and the value that would come from doing it. If the risk is low, the organization may choose not to invest. If the risk is high, stronger controls become easier to justify.

Heller is also a strong believer in minimum viable products and incremental progress. Organizations need something that works, reduces risk, and creates value. From there, the capability can mature over time.

One area where this becomes particularly important is metadata. Organizations may have data. They may even have a lot of it. But without definitions, ownership, lineage, and context, people struggle to trust what they’re seeing.

Generative AI can help enrich definitions, improve documentation, and make information easier to consume. But it still depends on a strong foundation underneath.

Read more: How to Create Data Management Processes That Enable Reliable, Repeatable Decisions

How to build a data governance program that enables business

Many data governance programs begin with good intentions. Leaders create councils, define roles, establish policies, and build approval processes. The assumption is that if the framework exists, better data practices will follow.

Heller has seen a different reality. In his experience, governance struggles when it becomes disconnected from the business problems it is supposed to solve.

Instead of asking leaders to support a governance initiative, he would discuss the risks associated with a business process, a regulatory requirement, or a critical decision.

The focus shifted from governance for governance’s sake to governance as a business capability. That distinction became increasingly important as the organization grew and new regulations emerged.

The same philosophy shaped how he thought about controls. Many organizations try to implement every possible best practice. Heller’s objective was never to create the most comprehensive governance framework possible. It was to create enough governance to manage risk effectively while still allowing the business to move forward.

Everything beyond that could be improved over time. The result was a governance model that felt less like a policing function and more like an enablement function.

Rather than creating heavy approval gates, teams were allowed to move quickly while metrics, reporting, and controls provided visibility into outcomes.

Governance becomes much easier when people understand the risk of doing nothing.

How to drive adoption for data programs

The governance framework is established, ownership is defined, policies are approved, and the technology is in place. Yet behavior often remains unchanged. People continue doing things the way they always have.

For Heller, this is where many data programs begin to lose momentum. The common instinct is to create mandates, assign responsibilities, define requirements, and escalate non-compliance.

Heller believes that it often creates resistance rather than engagement. Instead, he starts by asking questions:

• Do we have critical data?
• Who depends on it?
• What happens when it’s wrong?
• Who owns the risk?

When leaders begin talking about shared risks and shared outcomes, the discussion changes. Governance stops feeling like additional work and starts feeling like a solution to a business problem.

Throughout his career, Heller focused on building capabilities that solved real problems: better data quality, reduced risk, faster decision-making, and stronger controls. When teams experienced those benefits firsthand, participation often followed naturally.

Also, Heller disagrees with the idea that a successful program requires universal adoption. If 10% of the organization is actively engaged, but supports 90% of the company’s most critical processes, that’s not failure but impact.

Executive sponsorship also plays an important role. According to Heller, the most effective executives help remove barriers. They align priorities, resolve conflicts between teams, and reinforce why the effort matters.

A vital lesson from Heller’s experience is knowing where to draw the line. Regulatory obligations must be enforced and everything else should compete on value.

How to stay grounded through technology waves

The data industry has a habit of reinventing itself. Every few years, a new term captures the industry’s attention: big data, data lakes, data mesh, data fabric, observability, generative AI, agentic.

After more than a decade as a CDO, Heller has reached a somewhat counterintuitive conclusion: “The underlying problems in data management rarely change.”

The observation comes from experience. Long before generative AI, organizations already were dealing with issues around trust, governance, ownership, documentation, and accountability. The technologies were different, but the challenges were familiar.

Heller points to a large-scale migration project he worked on during the Y2K era. The work would sound familiar to any modern data leader. Teams still needed to document requirements, map data, validate quality, test outcomes, and manage risk.

“Organizations tend to anchor too heavily on tools and labels rather than the processes they are meant to enable.”

That mindset has helped Heller avoid getting distracted by technology hype. He doesn’t dismiss new technologies. He evaluates them through a different lens:

• What business problem does this solve?
• What risk does it reduce?
• What capability does it improve?

If the answer is meaningful, the technology deserves attention.

One example is the growing conversation around data products. Heller believes the concept is valuable, but not particularly new. Organizations have always benefited from curated, trusted data assets that reduce friction for analysts and business users.

What AI changes is the urgency. Suddenly, data quality, metadata, and governance are no longer operational concerns. They become AI concerns.

Who or what is a CDO today?

One of the questions we posed was straightforward: What makes a successful Chief Data Officer?

Heller does not believe there is a single blueprint. Over the years, he has worked with data leaders from a variety of backgrounds. What mattered was their understanding of how organizations work.

Throughout our conversations, Heller repeatedly described data management as a means to an end rather than an end in itself. That’s why he believes the role sits at the intersection of people, process, and technology.

A CDO spends part of their time with engineers and architects. Another part with business leaders. Another part with risk, compliance, legal, and audit teams.

Success often depends on being able to move comfortably between those worlds. “If you can’t talk a technical language to your technology partners, and business language to business partners, it’s hard.”

Further, Heller describes data leadership as “fireproofing.” The best data programs prevent problems before they occur. They reduce risk, improve controls, increase trust, and enable better decisions. When they work, nothing dramatic happens. And that’s precisely the challenge.

The burden then falls on the CDO to connect the work to outcomes, such as reduced losses, improved efficiency, better regulatory readiness, and stronger business performance.

Technology along with the scope of the role will continue to evolve. Similarly, AI will continue to create new expectations and new debates around ownership. But for Heller, the leaders who last are the ones who remain focused on business impact. Everything else is secondary.