The Death and Rebirth of Data (Part 3)

In Part 1 of this series, "The Day They Merged Data and Engineering Is the Day the Data Industry Died," explored how the fusion of data and engineering shifted focus away from business outcomes and toward technology for technology's sake.

Part 2, "The Death and Rebirth of Data," examined how the industry's obsession with pipelines, platforms, and plumbing left most organizations drowning in complexity rather than insight.

This third installment digs deeper into when and why data stopped belonging to the business, and how that loss of ownership created the "data priesthood" so many organizations now depend on. Reclaiming value from data doesn't start with new tools; it starts with returning ownership, context, and trust to business domains. Drawing from research by Gartner, McKinsey, Forrester, Deloitte, AWS, and BARC Germany, this piece explores how the next era of data and AI, will be defined not by centralization, but by empowerment.

In Part 2, I explored how the data industry became obsessed with building elaborate plumbing, pipelines, orchestrators, and tools, while forgetting the very thing those systems were meant to deliver, the water of insight and understanding.

But there's a deeper problem underneath the pipes. Even if the plumbing worked perfectly, most organizations would still struggle to extract value. But why is that?

Because somewhere along the way, the data stopped belonging to the business.

Locked Out

The business stood outside its own data

• 1Value extraction lost. Business users could no longer turn data into decisions on their own terms.
• 2The data stopped belonging to them. Control shifted to those furthest from the question.
• 3Ownership was lost. Without ownership, context, trust, and momentum quietly evaporated.

I have been in the data industry for a long time, and I can remember in the early days of business intelligence, data lived close to its purpose. It was departmental and hadn't yet gone enterprise.

Finance owned the financial systems. Marketing owned campaign data. Operations owned supply-chain metrics. Business users understood what the data meant because they were the ones using it to make decisions. They were "close" to the data.

Then came the technology wave. As business intelligence scaled across enterprises, responsibility for "data" shifted to IT and, later, to data or data-engineering teams. Suddenly, those who used the data were no longer the ones who controlled it.

What had been a business asset quietly became a technical domain. Ownership slipped from business leaders to technologists, and with it understanding, and the ability to move fast, ask new questions, and act with confidence. The "ownership" was no longer "close" to the data.

As I wrote in Part 1, this was the moment the data-engineering merge created an unintended divide, separating business context from technical capability.

Gartner's 2024 Data & Analytics Impact Report reinforces this, finding that less than 45 percent of business leaders trust their organization's data, largely due to the absence of domain-level ownership.

McKinsey's State of Data 2023 echoes this, noting that when responsibility remains siloed within IT, decision latency can increase by up to 60 percent.

The Keepers of the Data

Robed in expertise, the few stand between the business and its own data.

70%

of central team time spent on ad-hoc requests
Forrester, 2023

<45%

of business leaders trust their organization's data
Gartner, 2024

+60%

decision latency when data sits inside IT
McKinsey, 2023

Many companies now operate what can only be described as a data priesthood, a centralized group of experts who alone can interpret and manipulate the sacred data.

Need a metric? Submit a ticket. Need a new report? Wait for sprint planning. Need to fix a data definition? That's a backlog item for next quarter.

This system was created with good intentions, governance, prioritization, quality, and standardization, but it's come at a high price, dependency and disempowerment.

Forrester's Overcoming the Data Bottleneck (2023) found that centralized data teams spend up to 70 percent of their time on ad-hoc requests, leaving business users waiting for insights they should be able to access themselves.

The business users, once a curious explorer of their own data, has become a passive consumer of whatever the central team decides to deliver. And when decisions slow down or reports are delayed, the natural instinct isn't to question the model, it's to blame "the data team".

When ownership moved away from the business, context went with it. Engineers maintaining pipelines often have limited understanding of why a metric matters, and how it drives outcomes. Meanwhile the business stakeholders requesting data rarely grasp the technical constraints or assumptions behind how the data is delivered.

The result? A game of telephone across technical and business silos, where meaning is lost with each translation.

A single field like customer_value might be defined differently in marketing, finance, and product. And because no one "owns" the business definition end-to-end, inconsistencies multiply until trust collapses. That's how we ended up with the modern paradox. Despite record investment in data and technology, executives still don't believe or trust their data.

A Game of Telephone

Meaning frays a little with every translation.

• 1Engineers lost the "why". Pipelines are maintained without knowing which decisions they drive.
• 2The business lost the "how". Stakeholders rarely see the constraints and assumptions behind delivery.
• 3No one owns the meaning. customer_value drifts across marketing, finance and product until trust collapses.

Harvard Business Review's Why Data Culture Matters (2023) cites this exact phenomenon — inconsistent definitions leading to mistrust and misalignment.

According to BARC Germany's Data Culture Benchmark (2024), 72 percent of organizations cite "trust in data" as their biggest barrier to becoming data-driven.

In response, many organizations doubled down on centralization and control. They created data governance councils, committees, and approval processes to ensure consistency and quality.

But in doing so, they often made things worse.

A series of gates and approvals that slowed down progress rather than enabling it. It was designed to restrict, not enable.

Every change required approval.

Every new dataset required a steward. Every schema change needed a sign-off. Instead of empowering teams to manage their data responsibly, governance became a way to say "no" in a thousand different ways.

Gate After Gate

Governance built to enable became a corridor designed to delay.

• 1Every change required approval. Councils, committees, and sign-offs replaced momentum.
• 2Every dataset needed a steward. Stewardship became a queue, not a relationship.
• 3Designed to restrict, not enable. Governance became a way to say "no" in a thousand different ways.

As explored in Part 2, the industry confused control with quality. Good governance should enable safe usage, not prevent experimentation.

Deloitte's State of Data Governance (2024) reports that 84 percent of leaders believe governance slows innovation, while only 27 percent say it accelerates decision-making.

We forgot that good governance isn't about control, it's about clarity.

"Data governance is more than just adhering to regulations and compliance, it is the key to unlocking the full potential in other Data Management & Analytics capabilities." — Deloitte

That clarity empowers rather than constrains.

Reclaiming data's value starts with restoring ownership to where it belongs, the business. That doesn't mean dissolving the data team; it means reframing its purpose.

Here's how that shift looks in practice:

The central data function should be to enable business units, to manage and use their data, not control it at all.

Think of it as a platform team. They provide the infrastructure, tools, and governance guardrails so others can safely self-serve.

"True data democratization comes from domain ownership, not centralized pipelines." — Zhamak Dehghani, creator of Data Mesh

The success metric isn't "how many pipelines did we build", but "how many teams can now build their own".

Each business domain, marketing, finance, operations, should own the meaning, quality, and usage of its data. They are the ones that understand the context. The data team helps them publish and share those data products in a standardized way, but ownership stays local.

This is the essence of Data Mesh done right, not just a technical framework, but a cultural one.

BARC Germany's 2025 Study on Data Products calls this the "missing link between AI and operational business value." Ownership stays local; standards stay shared.

Rebuilding trust requires a shared semantic layer, a common language bridging technical definitions and business meaning. The marketing analyst and the data engineer should be able to describe "customer lifetime value" in the same terms, within their context.

AI tools and natural language interfaces can help here, translating between human questions and technical data structures, but the foundation is human alignment.

"Data reliability is a team sport — it only works when ownership and accountability are shared across business and data teams." — Barr Moses, Monte Carlo Data (Forbes Tech Council, 2023)

And as AWS notes, "A data-driven culture is only fully realized when data analytics skills are common across roles … and not exclusive to just data scientists."

When business users reclaim ownership, something powerful happens, curiosity returns. People start exploring again.

Questions that once took weeks to answer are resolved in minutes.

Teams stop asking, "Can I get this report?" and start asking, "What if we changed this decision?"

Data becomes alive again, not a static asset managed by technologists, but a living capability that fuels creativity, innovation, accountability, and enables decisions.

Ironically, this shift also liberates the data team. Instead of firefighting requests, they can focus on what they do best, building platforms, improving quality, advancing analytics, and enabling AI.

Everyone wins.

So, when did data stop belonging to the business?

The moment it became too technical for the business to touch.

• When dashboards turned into pipelines, and questions into Jira tickets.
• When we confused control with quality, and governance with gatekeeping.
• When the people who used to own the meaning of data lost access to it.

The rebirth of data begins by reversing that trend, giving ownership, understanding, and trust back to those who use it to make decisions every day.

As I wrote in Part 1, "The industry didn't die because we stopped building technology, it died because we forgot who we were building it for."

Remembering that truth, and building technology that brings data back home to the business, will be the rebirth of data.

• Gartner (2024). Data & Analytics Impact Report.
• McKinsey (2023). The State of Data.
• Forrester (2023). Overcoming the Data Bottleneck.
• Harvard Business Review (2023). Why Data Culture Matters.
• Deloitte (2024). State of Data Governance.
• Deloitte (2024). Data Governance and Management Insights.
• AWS (2024). Executive Insights — Becoming a Data-Driven Organization.
• BARC Germany (2024 & 2025). Data Culture Benchmark and Data Products Study.
• Dehghani, Zhamak (2022). Data Mesh: Delivering Data-Driven Value at Scale.
• Moses, Barr (2023). Forbes Tech Council: The New Rules of Data Reliability.