Data governance
8 min

Purview Data Governance Is Retiring: What to Carry Across, and Where the Meaning Should Live

Key takeaways

  • Purview Data Governance, meaning Microsoft Purview Unified Catalog and Data Map, is retiring. Microsoft says support ends on 12 November 2027 and names the OneLake catalog in Microsoft Fabric as its data governance solution.
  • Microsoft Purview itself continues as a security and compliance product. The retirement covers the governance capabilities only.
  • Microsoft's own architects grade the move outcome by outcome and call most of it partial. For glossary terms and critical data elements their instruction is to export and preserve terms and ownership.
  • A CSV export carries a name, a definition, an owner and an expert. It carries no approval record, no version history and no validation rule, so the receiving side has to supply those.
  • TopQuadrant holds the exported meaning as a governed record on open standards: a named expert confirms it, SHACL validates it, every past version stays answerable, and the record can be exported again.
In this article
Share

A data platform lead opened the Microsoft Purview Unified Catalog documentation on 13 November and found a new notice at the top of every page. The product their business glossary lives in is in maintenance mode, and support ends on a fixed date. The search for Purview alternatives now comes with a deadline.

Microsoft has named the destination: the OneLake catalog in Microsoft Fabric. So the question has changed. The old one was which catalog to buy instead. The new one is what to carry across, and where the business meaning should live once the catalog that held it is gone.

Purview Data Governance retires on 12 November 2027. Microsoft's replacement is the OneLake catalog, and Microsoft's own architects tell customers to export and preserve terms and ownership because full glossary parity is off the table. TopQuadrant's position is that the exported meaning should land in a governed, open-standard record that outlives any catalog, including the next one.

What did Microsoft actually announce?

The announcement went out on 12 November 2026 on the Microsoft Tech Community, and every Purview governance page on Microsoft Learn was updated the same day. The rows below are Microsoft's words.

ItemMicrosoft's wordsSource and date
What retires"Microsoft Purview Unified Catalog and classic Microsoft Purview data governance capabilities (formerly Azure Purview) are being retired." / "You can no longer create new accounts, and existing capabilities will remain in maintenance mode through November 12, 2027."Notice on every Purview governance page, Microsoft Learn, updated 12 November 2026
The date"Support for Purview Data Governance will end Nov. 12, 2027."Microsoft Tech Community announcement, 12 November 2026
The replacement"The OneLake catalog is now the single foundation for governing data across your entire estate."Same announcement
How it is paid for"it's included in Fabric capacity"Same announcement
What continues"Microsoft Purview stays focused on data security and compliance"Same announcement

Microsoft Purview continues as a security and compliance product, so nothing here says Purview is retiring. The retiring part is what Microsoft's blog calls Purview Data Governance: the Unified Catalog and the Data Map.

Why do teams search for Purview alternatives now?

Three reasons drive the search now, and they lead to different work.

The three reasons behind the search

  • A dated deadline. Maintenance mode through 12 November 2027, no new accounts, support ending the same day. Every term and data product in Unified Catalog has to be somewhere else by then.
  • A replacement whose parity Microsoft itself qualifies. The architects' transition post grades each outcome, and most are partial. A Fabric shop gets the catalog, and a list of what to keep outside it.
  • A question of where meaning lives. Terms and owners are the things Microsoft says to export. Exported to where, in what form, and with what still attached, is the open decision.

Fabric and a plan answer the first two. The third is this piece.

What is Microsoft Purview Unified Catalog genuinely good at?

A team that built on it built on a capable product, and the retirement changes only the timeline.

The Data Map scans wide. Microsoft's source table, dated 13 October 2026, lists 46 source rows, and the non-Microsoft entries include Snowflake, Google BigQuery, Oracle, Teradata and SAP (learn.microsoft.com, data-map-data-sources). Data quality runs on the data itself, with "Out of box rules to measure six industry standards data quality dimensions" (unified-catalog-data-quality, 13 October 2026).

Microsoft's glossary page, dated 13 October 2026, says: "Glossary terms in Microsoft Purview Unified Catalog are active values that provide context. They also apply policies that determine how you manage, govern, and make your data discoverable for use." (unified-catalog-glossary-terms). A term carries policy. A CSV row does not, and that is the first thing an export loses.

What do Microsoft's architects say carries across to the OneLake catalog?

The transition post is by Paul Singh and Rafia Aqil, Microsoft architects. It was published on 13 November 2026 and updated on 16 November 2026 (Microsoft Tech Community, 4573422). It maps each Purview outcome to Fabric and defines its own grading word: "Partial means part of the outcome is covered and part is not." Every row below quotes that post.

Purview outcomeMicrosoft's guidance, verbatimSource
Glossary and critical data elements"Export and preserve terms and ownership. Do not claim full glossary or CDE parity."Tech Community 4573422, 13 November 2026
Data products"Define the product contract, owner, health, lifecycle, and access process outside the catalog where needed."Same post
Data quality"Governance Health is not a rules, profiling, and scoring engine."Same post
Lineage"Retain another method for cross-system, unsupported-workload, or granular lineage."Same post
Workflows"Keep the request, approval, expiration, revocation, and evidence process where Fabric does not cover it end to end."Same post
Scanning"Data connectivity is not enterprise metadata scanning."Same post
Stewardship"Keep a responsibility matrix for stewardship duties without a dedicated role."Same post

Read the first row and the last two together. The terms and owners leave as an export. The approval process stays wherever you keep it. The steward becomes a row in a matrix. That is a definition with nobody attached to it by the system.

The OneLake catalog overview on Microsoft Learn is dated 26 August 2026, updated 29 September 2026, before the announcement (onelake-catalog-overview). It names "domains, endorsements, metadata scanning, policies, and governance insights". As of that date it names no glossary terms, data products or critical data elements. Fabric ships monthly, so read that page again before you plan around it.

What does "export and preserve terms and ownership" actually preserve?

Microsoft's glossary documentation shows what a term carries today. The page is dated 13 October 2026, updated 12 November 2026 (unified-catalog-glossary-terms-create-manage).

What the term carriesMicrosoft's documentation, verbatimWhat a file export keeps
Name, definition, owner, expertBulk import columns: "name (required)", "description (required)", "owners (required): Enter a valid email address", "experts"Kept, as text and email addresses
Who may define it"To create, edit, and manage glossary terms, you need the Data Steward role in the governance domain where the terms reside."Lost. The role belongs to the catalog
Change control"If the term is published, select Unpublish, and then select Edit to edit the name, definition, owners, or custom attributes."Lost. Draft, Published and Expired are catalog states
Attached policy"all the associated policies are automatically applied." (unified-catalog-glossary-terms)Lost. Policy stays with the catalog
Version historyThe pages we read document Draft, Published and Expired. They document no version history of a term.Nothing to export

So the export preserves the words and some email addresses. It preserves no record of who confirmed the definition, no rule that tests data against it, and no history of what it meant before. Those lived in the catalog around the term, and the catalog is what is being retired.

In our post Metadata Management vs Data Governance: What Each One Decides we state that ownership has to live on the record itself. An owner in column C of a CSV is a document.

What does TopQuadrant do with the exported meaning?

Our position, with the same weight as Microsoft's evidence above and open to the same test.

Microsoft's announcement says: "Fabric also integrates with other 3rd party providers such as Atlan, Alation, Collibra and Informatica so customers can decide what works best for their data governance requirements." TopQuadrant is absent from that sentence, and no live TopQuadrant page describes a Purview or OneLake catalog connector. We claim neither. What we answer is the other half of the architects' instruction: the receiving side.

What the receiving side holds that the export does not

  • The term becomes a governed object. From our site: "Your experts model the concepts, relationships, rules and processes your business actually runs on, approve them, and govern them in one place." The owner, the scope and the confirmation are held as data.
  • Proposing and confirming are separate steps. "AI proposes. Your experts confirm. Inference gets you started, and agreement is what makes it true." An exported CSV is a proposal. The owner in column C becomes the named confirmer on the record, with a date.
  • The definition carries a test. TQ Data Foundation validates with SHACL constraints. A definition with a rule attached can be checked against the data on every load. That is the "rules, profiling, and scoring engine" Microsoft says Governance Health is not, applied to the meaning itself.
  • Every past version stays answerable. From our reference data page: "Point-in-time queries. Ask for the list as of any past date and get the state that was live then." The audit question after a migration is what a term meant on the day you moved.
  • The record is on open standards. TQ Data Foundation is "Built entirely on open knowledge-graph standards · RDF · OWL · SHACL". TopQuadrant helped author SHACL and the RDF specifications. The thing Microsoft told you to export arrives in a form built to be exported again: "Your context is yours: export it and go if we stop earning the work."
  • The agreed record is served live. Approved context reaches applications and agents through MCP and SDKs. "A steward's change does not need a deployment in every downstream system to take effect."
  • The start is a model. We maintain 300 reusable standards-based models, so a migrated glossary lands in governed structure on day one.

On the catalog, said once

Our usual line is that your catalog stays and syncs from the governed record. The live wording names three catalogs: "Collibra, Informatica, and Alation reflect the same values and effective dates the stewards maintain." The catalog is moving. The meaning should not have to. Fabric becomes the catalog your people search, and the definitions it shows come from a record with a name on it. In our post What Is Metadata Management? we state that a catalog is organized around assets and meaning around concepts. That is why one can be swapped while the other is kept.

Where we do not compete: scanning across 46 source types, pipeline lineage, data quality on the data, and protection. Those stay with Fabric and Microsoft Purview, and a Fabric shop gets the catalog basics within capacity it already pays for. We cannot match "included". We hold the meaning those tools display.

How should you decide what to carry across?

Four questions, in order.

Four questions that settle the migration

  • What is your real cut-off? Support ends on 12 November 2027. Your freeze date is earlier, because every edit means unpublishing a term and publishing it again.
  • Which of the seven outcomes do you actually use? Score your own estate against the architects' table. A regulated team has all seven.
  • Where will the terms and owners live, and in what format? In a CSV, the owner is an email address and the definition is a string. In a governed record on open standards, both are data with a date.
  • Who confirms a definition after the move, and where is it recorded? Microsoft's guidance keeps stewardship in a responsibility matrix without a dedicated role. Decide where the confirmation itself gets written down.

Our post Metadata Management Tools for Governance and AI Readiness covers the wider market. Our What Is Data Governance for AI? post covers why the confirmation record matters once the reader is a model.

Where does TopQuadrant fit?

As the place the exported meaning lands. Microsoft has decided the catalog. TopQuadrant holds the terms, the owners, the rules and the history those terms need to stay true after the move.

If your Purview estate is scans and lineage, Fabric and a plan are enough. If it is a glossary your business agreed on, the export is the fragile step. Carry the words into TQ Data Foundation. Let the named owner confirm them. Attach the rule. Keep every version. The next catalog retirement then costs you a connector, and never the meaning.

Give your AI the context it's been missing

See how the TQ Data Foundation turns your enterprise knowledge into trusted, Al-ready context.