Data Governance at a Crossroads

A Longstanding Discomfort, Poorly Identified Causes

Data governance remains, even today, one of the most debated topics in enterprise digital transformation. And this discomfort is nothing new.

For years, organizations have launched ambitious initiatives, mobilized teams, invested in tools, structured roles… only to encounter deep difficulties: limited business adoption, operational heaviness, conflicts of responsibility, difficulty demonstrating value creation, and sometimes even a gradual rejection of the initiative itself.

The causes are most often sought within the business units, IT, organizational culture, tooling, or lack of maturity. And the responses logically follow that diagnosis: new cataloging tools, data literacy programs, change management initiatives, organizational acculturation efforts…

Often serious and sometimes costly treatments — but applied to the wrong diagnosis. Because when the very foundations of the approach are poorly framed, no amount of change management will produce the expected adoption. The symptoms are treated without addressing the root cause.

What the Major Approaches Proposed — and What They Imply

DAMA — Governance as the Exercise of Authority

DAMA International defines data governance as the exercise of authority, control, and shared decision-making over the management of data assets.

What this implies: data is an asset, and governing means exercising authority over that asset. The natural structure that follows is hierarchical — roles, rights, and decision-making bodies. Governance becomes a control function, operating with a logic similar to that of finance or legal departments.

What this leaves unanswered: why is this authority exercised? In service of what? The purpose remains implicit. And in organizations where authority is diffuse or contested — which is to say, most organizations — this framework produces conflicts of legitimacy more often than it resolves them.

DGI — Governance as a System of Rights

The Data Governance Institute structures governance around decision rights: who can do what with which data, under which circumstances, and according to which methods.

What this implies: governance is first and foremost a legal and organizational mechanism. It defines boundaries, assigns rights, and clarifies formal responsibilities. This is a real step forward compared with DAMA — the question of who decides is addressed more explicitly.

What this leaves unanswered: rights are defined on data, not on the decisions that data is meant to support. One may know who can modify an attribute, but rarely why that attribute exists and which decision it conditions. The framework remains centered on data as a legal object, not as an instrument for action.

Ladley — Governance as Behavioral Transformation

John Ladley shifts the focus toward people: governing data means organizing behaviors, building a culture, and engaging organizations in sustainable transformation.

What this implies: resistance to change is identified as the real obstacle. The solution is therefore cultural — acculturation, adoption, change management. This perspective resonated strongly across many organizations and partly explains why data literacy and change management programs have been so widely mobilized.

What this leaves unanswered: changing behaviors around what, exactly? If the purpose is unclear, cultural transformation becomes directionless. Adoption is created without orientation. And when results fail to materialize, fatigue sets in all the more quickly because the human investment has been substantial.

Seiner — Non-Invasive Governance

Robert Seiner proposes a deliberately minimalist approach: governance must fit into existing practices without disrupting them. It is non-invasive, or it fails.

What this implies: the diagnosis is one of friction. Previous initiatives failed because they were too heavy, too constraining, too disconnected from the daily reality of teams. The solution is therefore lightness — building on what already exists, formalizing without imposing.

What this leaves unanswered: non-invasiveness is a deployment constraint, not a purpose. It explains how to deploy, not why. And in situations where deep changes are required — and they often are — it can become as much a limitation as an enabler.

Deployment Models and Data Mesh — Governance as an Architectural Question

Another family of approaches sought the solution in organizational structure: centralized, federated, or hybrid governance. More recently, Data Mesh crystallized this trend by proposing domain-based decentralization as a response to the failures of centralized governance.

What this implies: the problem is structural. If governance fails, it is because it is poorly organized, too centralized, too disconnected from data producers and consumers. The solution is therefore architectural — redistributing responsibilities and bringing governance closer to business domains.

What this leaves unanswered: a deployment model does not define a purpose. Entire organizations have launched major transformations toward Data Mesh hoping decentralization would solve the adoption and accountability issues their previous governance frameworks had failed to address. Often, they encountered the same disappointments — simply relocated to the domain level.

What These Divergences Reveal

These approaches are not incompatible by accident. Each starts from a different diagnosis — authority, rights, culture, friction, architecture — because they never shared the same answer to the most fundamental question: what is data governance actually for?

This is not a minor detail. It is the source of all the divergences. And as long as this question remains unanswered explicitly, every organization will continue adopting the framework that resonates most strongly with the symptom it perceives — without being able to assess whether that framework truly brings it closer to the problem it is trying to solve.

The work ahead is considerable. Not because these approaches are without value — each has produced real advances, useful tools, and practices that have helped organizations progress. But because they have also left deep traces: teams trained within a certain logic, tools deployed around a particular vision, organizational cultures built upon specific beliefs.

Deconstructing all this — without discarding what works, without alienating those who invested years into these approaches, without losing those who have already disengaged — is an entirely different challenge from merely proposing a new framework. That may well be the most underestimated part of what still lies ahead.

Paths from Which Organizations Must Now Return

These paths were pursued seriously, often with conviction, and always with significant resources. Walking back from them is neither simple nor painless. Teams have been structured, tools deployed, operational models built. Questioning the foundations also means questioning years of work — and the people who carried it.

And then there are all those who were lost along the way.

Those who never truly understood where they were being led and merely pretended to engage. Those who understood, tried, and encountered resistance they had no means to overcome. Those who became exhausted defending an approach whose value they could no longer demonstrate. And finally, those who disengaged entirely — and who will be difficult to win back because they have already invested heavily.

How can they be brought back? Certainly not with a new framework, a new acronym, or another promise of transformation. The distrust that has emerged is real. It will only fade under one condition: that the proposed approach finally starts from a purpose that is clear, understandable, and directly connected to the issues organizations genuinely recognize as their own.

That may well be the real challenge of the next stage — not convincing those who never tried, but restoring meaning for those who did and no longer believe.

A Definition to Anchor the Approach

Data governance is an approach that consists in using data to frame and validate business decisions, whether those decisions are made by humans or delegated to systems, partners, service providers, or machines.

At first glance, this definition may appear close to previous ones. Yet it is far more than a simple reformulation. It introduces several fundamental shifts.

Shift 1: Data Is No Longer the Object — It Is the Instrument

In the definitions proposed by DAMA, DGI, or Ladley, data is what is governed. It is the central object around which roles, policies, and processes are organized.

Here, data becomes what governance operates through. It is the instrument of a governance act whose true object is the decision itself. This reversal is not semantic: it radically changes what is measured, what is built, and what organizations seek to demonstrate.

Shift 2: Business Decisions as the Fundamental Unit

Where traditional frameworks organize governance around data domains, entities, attributes, or flows, this definition proposes starting from business decisions: a credit decision, a pricing decision, a hiring decision, a regulatory decision.

Each decision has its own structure: actors, required data, applicable rules, timelines, responsibilities, and conditions for contestability. Governing data means ensuring that all these conditions are reliably, traceably, and auditable fulfilled.

This approach makes governance immediately understandable to business teams — not as an abstract constraint, but as a response to concrete decision-making risks they already recognize.

Shift 3: Delegation as a Central Dimension

This is perhaps the newest — and most urgent — dimension. The definition is not limited to decisions made by humans. It explicitly integrates decisions delegated to systems, partners, service providers, or machines.

Traditional definitions were built at a time when human decision-making was the dominant model. They are not equipped to address today’s challenges: scoring algorithms, recommendation engines, automated detection systems, generative AI agents operating with near autonomy.

Yet delegating a decision to a machine does not eliminate responsibility — it shifts and complicates it. Who is accountable when an algorithm makes a poor credit decision? When an automated system incorrectly applies a compliance rule? When an AI agent communicates on behalf of the organization without human supervision?

Data governance must answer these questions. Traditional frameworks, by remaining silent on delegation, leave organizations without guidance in the face of rapidly growing decision-related risks.

Shift 4: Trust as a Measurable Purpose

Finally, this definition implicitly shifts the measure of success. Success is no longer the number of defined rules, covered domains, or appointed stewards. It is the level of trust the organization can place in its decisions: their traceability, auditability, compliance, and consistency over time.

A trust that must be demonstrable — to regulators, partners, customers, employees, and executives who increasingly bear responsibility for decisions that are both more complex and more automated.

What This Changes for Existing Initiatives

Accepting this definition means accepting to look differently at what is already in place. Not to destroy everything, but to assess each existing governance component against a simple question: does it contribute to securing a real decision, carried by identified responsibilities, under defined conditions of trust?

Some deliverables will withstand that assessment — and this will prove they were genuinely anchored in business reality. Others will reveal that they were produced for governance itself, without any explicit connection to a concrete decision or responsibility. This is not a judgment — it is a starting point.

For ongoing initiatives, this concretely implies:

· Requalifying priorities around the most critical business decisions, rather than seeking exhaustive coverage of data domains

· Rearticulating roles — data owners, stewards, coordinators — around real decision-making responsibilities, not abstract data perimeters

· Redefining success indicators: no longer the number of documented rules or certified domains, but the ability to demonstrate that a given decision is better framed, more traceable, and more auditable than before

· Explicitly integrating delegation — to systems, partners, and service providers — as a governance object in its own right, with its own requirements for traceability and accountability

This transformation will not happen overnight. But it can begin now, without waiting for a complete overhaul — simply by asking, for every existing initiative, what its actual decision-making purpose is.

What Comes Next

This definition is a starting point, not a conclusion. It opens questions that the next articles in this series will explore: how can organizations identify the business decisions that require strengthened governance? How should human responsibilities and delegation to systems be articulated? How can the trust created by governance be concretely demonstrated?

The keys already exist. They are emerging from the practices of those who have managed to move beyond the current dead ends — and they deserve to be shared.

Next article:What Is a Business Decision in the Context of Data Governance? Anatomy of an Object Left in the Shadows for Too Long.

© 2025. All rights reserved.

Mentions légales

Politique de confidentialité

Nous rejoindre