OneLake Catalog vs Microsoft Purview: What Does What

The OneLake catalog and Microsoft Purview both live in the Microsoft governance world, both talk about data discovery and trust, and both promise to help you find and govern data. So the question "which one do we need" is fair and common — and answering it wrong means either duplicated effort or a gap in governance.

The short answer is that they are complementary, not alternatives. They operate at different scopes and do different jobs, and most organisations of any size use both. The useful thing is to be clear on where each fits so you deploy them deliberately rather than assuming one covers the other.

Getting this right matters more now that AI agents ground on your data, because both tools play a part in making the data agents reason over trustworthy — and a gap between them is a gap the agents inherit.

The OneLake catalog is the discovery and governance hub inside Fabric. Its Explore experience helps anyone find the Fabric data they can access — search, tags, descriptions, endorsements — so people locate the trusted dataset instead of rebuilding it. Its Govern experience shows the governance posture of the Fabric data you own and recommends actions to improve it: add descriptions, apply endorsements, address sensitivity.

Its scope is Fabric. It covers the full range of Fabric item types — lakehouses, warehouses, semantic models, reports, dashboards — and it is embedded where Fabric users work, in Teams, Excel and Copilot Studio, and queryable through public APIs, the Fabric CLI and the Fabric MCP. It is the day-to-day catalog for a Fabric estate.

Think of it as the tool that makes your Fabric data discoverable, understandable and endorsed, and gives data owners a live view of how well governed their Fabric items are. It is where in-Fabric governance happens.

Microsoft Purview is enterprise data governance and compliance across your whole estate, not just Fabric. Its remit is broader: sensitivity labelling and classification, data-loss prevention, compliance and risk, and data lineage and cataloguing across many sources — on-premises databases, other clouds, SaaS applications, as well as Fabric.

Where the OneLake catalog governs Fabric data specifically, Purview provides the enterprise-wide layer: a consistent sensitivity taxonomy applied everywhere, compliance controls that span sources, and visibility across data that lives well outside Fabric. It is the tool a governance or compliance function uses to manage data risk across the organisation.

So Purview answers "how is data classified and governed across everything we have", while the OneLake catalog answers "how do people find and govern the data inside Fabric". Different questions, different scopes.

The two connect rather than compete. Purview sensitivity labels and classification flow onto Fabric data, so the enterprise taxonomy applies inside Fabric, and the OneLake catalog surfaces that sensitivity in its discovery and posture views. You define sensitivity and compliance once in Purview at enterprise scope, and the OneLake catalog is where Fabric users experience and act on it day to day.

A practical division: use the OneLake catalog for in-Fabric discovery, endorsement and the governance posture of Fabric items; use Purview for enterprise sensitivity classification, compliance, DLP and cross-source lineage; and connect them so Fabric data inherits enterprise classification rather than being governed in isolation. Neither replaces the other, and using only one leaves a gap — Purview alone misses the in-Fabric discovery experience, the OneLake catalog alone misses the enterprise-wide compliance layer.

For AI grounding this pairing matters: agents reason from endorsed, classified Fabric data, where the OneLake catalog provides the endorsement and discoverability and Purview provides the classification. Together they are what make the data an agent grounds on trustworthy and compliant.

If your estate is centred on Fabric and the immediate need is helping people find trusted data and improving the governance posture of Fabric items, start with the OneLake catalog — it is part of Fabric and needs curation more than procurement. If you need enterprise-wide sensitivity classification, compliance controls and lineage across sources beyond Fabric, that is Purview.

For most organisations of any scale the answer is both, connected: Purview for the enterprise governance and sensitivity layer, the OneLake catalog for the in-Fabric discovery and posture, with classification flowing between them. Deploy them to their strengths rather than stretching one to cover the other.

And whichever you lead with, remember why it matters now: discoverable, endorsed, classified data is what makes both self-service analytics and AI agents trustworthy. Governance stopped being a back-office compliance exercise the moment agents started grounding on your data — it became the thing that decides whether the answers can be trusted.