Data Governance Solutions

Forrester's 2014 evaluation of this market was called Data Governance Tools, and its stated ambition was that vendors were preparing what the report termed data governance 2.0 management tooling. The audience was enterprise architecture professionals.

Eleven years later the same market is described as the control plane for trust, agility, and scale, and Forrester's position is that data governance has outgrown its compliance roots.

Between those two framings sits a decade in which data governance software was bought enthusiastically, deployed partially, and used less than anyone intended. The interesting question is not why the language changed. It is what changed underneath that finally made the software matter.

What the software actually does

A data governance platform maintains an authoritative account of an organisation's data. Six functions, in rough order of how early they appear in a deployment.

Cataloguing, meaning a searchable inventory of what data exists and where it lives. Business glossary, meaning agreed definitions of terms the organisation uses, and the mapping from those terms to actual columns in actual systems. Lineage, tracing where a piece of data came from, what transformed it, and what depends on it downstream. Classification, identifying which data is sensitive, regulated, or restricted. Policy management, expressing rules about who can use what for which purpose. And stewardship workflow, meaning the human process by which someone owns a data domain and resolves questions about it.

Forrester's current criteria set spans data discovery, business glossary, catalogue, semantics, lineage acquisition and visualisation, data observability, and AI governance, which is a fair description of how the scope has broadened.

Lineage deserves particular attention because it is the capability that has become load-bearing. Knowing that a number in a board report descends from three joins across two systems, one of which is a manually maintained spreadsheet, is the difference between a figure you can defend and a figure you hope is right.

Why it kept underdelivering

The category has a long history of expensive deployments that produced a well-populated catalogue nobody opened, and the reasons are structural rather than accidental.

Governance was framed as compliance. That framing releases budget, because regulation makes it mandatory, and it poisons adoption, because it positions the function as an obligation rather than a service. A platform bought to satisfy an auditor gets designed for the auditor.

Governance was staffed as a gatekeeper. The stewardship model asked a small central team to review, approve, and control access to data that a much larger population wanted to use. Analysts who found the process slow simply routed around it, extracting data into spreadsheets where no governance applied at all. The catalogue documented the governed estate while the actual work happened elsewhere.

And the value was diffuse. A well-governed data estate prevents problems that never happen, which is the hardest kind of benefit to defend in a budget review against a project that promises revenue.

Add these together and you get the pattern the industry knows well: a substantial licence, an eighteen-month implementation, a catalogue populated by a consultancy, and a slow decay as the estate changes and nobody maintains the documentation.

What made it load-bearing

Generative AI changed the argument in a way that a decade of governance advocacy could not.

The mechanism is direct. An AI system answering questions about your business retrieves from your data. If the retrieval surfaces an outdated policy, the model states outdated policy as fact. If it surfaces a metric defined three different ways in three systems, it picks one and presents it confidently. If it retrieves data an employee should not see, the information reaches them through the answer rather than through a permission error.

Every one of those failures is a governance failure appearing as an AI failure, and it is far more visible than the governance failures that preceded it. A badly governed data warehouse produced disagreements in meetings. A badly governed AI assistant produces a fluent, cited, confidently wrong answer at scale, to anyone who asks.

That visibility is what converted governance from a cost centre into a prerequisite. The organisations that had done the unglamorous work of definitions, lineage, and classification found their AI programmes worked. The organisations that had not discovered that the model was not the problem.

Forrester's framing captures the reversal: governance is no longer perceived as a business blocker, and in an AI-fuelled and data-saturated enterprise it functions as the control layer for trust and scale rather than as a compliance exercise.

Inside The Forrester Wave: Data Governance Solutions, Q3 2025

The evaluation scored thirteen providers across current offering, strategy, and customer feedback, with criteria covering data discovery, business glossary, catalogue, semantics, lineage, observability, and AI governance.

Atlan placed as a Leader and was recognised as a Customer Favorite, described by Forrester as a top choice for organisations wanting a modern, AI-native governance platform blending intelligent automation with deep integration and broad accessibility. Its cited strengths include policy management, stewardship, and collaborative governance, alongside fast deployment and governance embedded into daily workflows. Forrester specifically credited superior adoption across customer organisations, attributing it to ease of use and a commitment to data literacy.

Adoption as a differentiator is the finding that matters in this category. Given the history described above, a platform people actually open is doing something the category has historically failed at.

Alation also placed as a Leader with the highest possible score across eleven criteria including vision, roadmap, and adoption. Forrester characterised its strategy as shifting governance from passive documentation to intelligent, agentic workflows, with a vision of embedding governance into daily work and aligning metadata with business outcomes, supported by early investment in intelligent policy suggestions, automated stewardship triggers, and context-aware recommendations. Reference customers praised the interface and integration flexibility while identifying data observability and native data quality tooling as areas needing continued investment.

Collibra placed as a Leader as well, and its presence is notable for a different reason.

The one name that survived

Forrester's Q2 2014 evaluation, authored by Henry Peyret with Michele Goetz, scored ten providers against twenty five criteria: Adaptive, ASG Software Solutions, Collibra, Global IDs, IBM, Informatica, Information Builders, SAP, SAS Institute, and Trillium Software.

Collibra is the only name from that list that remains a standalone Leader in the current evaluation. The large platform vendors persist as companies but not as the reference points of this category. Several of the specialists no longer exist independently.

Set against that, both other Leaders in the current Wave are companies that did not appear in 2014 at all. Atlan and Alation are products of the modern data stack era, built around cloud warehouses, distributed teams, and self-service rather than around centralised enterprise architecture.

That is a fairly complete generational replacement, and it tracks the shift in what the software is for. The 2014 products were designed to help a central architecture function document and control an estate. The current Leaders are designed to help a distributed population of practitioners find, trust, and use data, with governance applied along the way rather than at a gate.

The consolidation pressure is real too. Recent acquisition activity in this space reflects large technology providers treating data governance as foundational to AI readiness, which is a polite way of saying the independents are attractive targets precisely because the category became strategically important.

Provenance becomes the product

The capability worth weighting most heavily in a current evaluation is lineage, and specifically lineage that survives contact with AI systems.

Here is why. When a human analyst produces a number, the organisation can ask them where it came from. When a model produces an answer, that question has to be answerable by the infrastructure, because there is nobody to ask.

Regulatory attention is arriving at exactly this point. Emerging AI regulation across several jurisdictions places obligations around training data provenance, documentation of data sources, and the ability to explain how an automated decision was reached. Those obligations do not land on the model. They land on the data pipeline feeding it, which is what a lineage graph describes.

The practical consequence is that lineage stops being a diagram someone drew during implementation and becomes something that must be automatically acquired, continuously current, and complete enough to answer a question posed by an auditor about a specific output.

Forrester scoring lineage acquisition separately from lineage visualisation is a meaningful distinction here. Visualisation is a diagram. Acquisition is whether the system can work out the lineage itself from the systems it connects to, without a human documenting it. Only the second one stays accurate.

Who owns this now

The unresolved question in this category is organisational rather than technical, and it determines whether a deployment succeeds more than the vendor choice does.

Data governance historically sat with a chief data officer or under enterprise architecture, with a small central team and a mandate that exceeded its authority. That model produced the gatekeeper dynamic that limited adoption.

The AI-era model implied by the current Leaders is different. Governance is embedded in the tools people already use, policy is suggested and enforced automatically rather than reviewed manually, and stewardship is distributed to the domains that actually understand their data, with the central function setting standards rather than approving requests.

That is a better model and it requires something most organisations have not done, which is deciding that domain teams own their data and are accountable for its quality. Buying a platform designed for distributed stewardship and operating it with a central gatekeeper team reproduces the old failure with better software.

The organisations getting value from this category right now are the ones where the AI programme created enough urgency to settle the ownership question that a decade of compliance argument never could. That is an unglamorous reason for a market to finally work, and it is the reason this one is.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.