A claims analyst looking for paid claims in Fabric might find a warehouse table, a Power BI semantic model, and three reports with nearly the same name. The analyst needs to know which one Finance uses and whether the paid amount includes reversals. If the number goes into a reserve review, someone may also need to trace it back to the source and show who approved access. All of that work falls under data governance, and the catalog has to support more than the initial search.
On November 12, 2026, Microsoft announced the retirement of Purview Data Governance and named Microsoft Fabric, with OneLake Catalog, as its data governance platform. Support ends November 12, 2027. Microsoft’s Partner Center notice puts Unified Catalog, Data Map, Workflows, and the older Azure Purview governance offerings in scope. Microsoft has closed new governance account creation and put existing capabilities in maintenance mode through the end of support.
Microsoft Purview continues to handle data security and compliance. In Fabric, that includes sensitivity labels, data loss prevention, and audit. A steward who depended on a Data Map scan to identify sensitive fields in an upstream system still has to account for that scan in the transition. The security policy and the source scan do different jobs.
Where OneLake Catalog fits
I like the direction for organizations that already do much of their analytical work in Fabric. In OneLake Catalog, the analyst can use Explore to find the certified claims model, identify its owner, and see the lineage Fabric captures. If the model is configured to be discoverable, the analyst can request access from the catalog. The team that owns the model can use Govern to check curation and sensitivity label coverage, and workspace administrators can use Secure to review permissions and manage supported OneLake security roles. The people making those decisions work with the same items the analyst finds.
That proximity matters when someone uses the model outside the Fabric portal. An analyst can discover it from Excel. If the team makes it available through a Fabric data agent, the agent honors the user’s permissions on the underlying data. Finance still owns the definition of paid claims. A clear description and a certified model give people a better chance of finding the measure Finance actually uses.
OneLake also reaches selected data outside Fabric. A shortcut can reference files in Amazon S3 from a Fabric lakehouse. Metadata mirroring can surface supported Azure Databricks Unity Catalog tables while reading the data in place. Database mirroring, as with Azure SQL Database, continuously replicates data into OneLake. Those are different ways to bring external data into Fabric’s working environment. With an external shortcut that uses a configured connection identity, for example, the source authorizes that identity and OneLake controls which Fabric users can read through the shortcut. That detail matters if the analyst also has a direct way to reach the source.
The jobs that need a replacement
A team that used Unified Catalog to find Fabric assets has a straightforward starting point. A steward who used it to maintain a business glossary, approve access to a data product, and report on quality across several platforms has a broader migration. Here is how I would map that work against the capabilities Microsoft documents today.
| Purview capability | What Fabric covers today | What the transition needs to cover |
|---|---|---|
| Discovery and domains | OneLake Catalog searches Fabric items and supported tables. Owners, tags, and endorsements provide context, and Fabric domains group workspaces by business area. | Carry over who owns and stewards each business domain, including assets managed outside Fabric. |
| Data Map scanning | Fabric offers tenant metadata scanning for Fabric items. Shortcuts and mirroring make selected external data or catalogs visible in Fabric. | Replace scheduled scans of external systems where teams rely on Data Map’s source-specific inventory, schema, and classification coverage. |
| Lineage | The catalog shows dependencies for supported Fabric items. The workspace lineage view can show external upstream sources one step away. | Preserve the path through external transformations, including column-level lineage where a report or control depends on it. |
| Business glossary and data products | Fabric item descriptions, endorsements, and semantic models carry business context into discovery and use. | Rehome governed glossary terms, relationships to critical fields, and curated data products that group assets across platforms. |
| Data quality and health | Govern tracks catalog and security posture. Fabric can apply row-level quality constraints during materialized lake view refreshes. | Replace the profiling, recurring rule runs, and quality scores reported across assets, products, and domains. Preserve the history teams use to monitor changes. |
| Access and publication workflows | Fabric supports access requests for discoverable semantic models and provides controls for reviewing and granting Fabric permissions. | Replace product access and publication workflows, including approver routing, terms-of-use attestations where required, and approvals to publish glossary terms. |
Take the claims example. The Fabric lineage view can show the warehouse and semantic model behind a report. If a Snowflake transformation assigns the reversal flag, Finance needs to be able to follow the number through that transformation as well. Fabric can consume a supported Snowflake output through a shortcut. Tracing how the reversal flag was calculated requires lineage from the Snowflake transformation itself. For a reserve review, Finance needs to be able to follow both parts of that path.
Quality is a similar case. A Fabric engineering team could implement a rule in a materialized lake view to keep records with a missing payment identifier out of the curated table. The steward’s job includes tracking whether the rate of missing identifiers is rising and finding which upstream system introduced the problem. If the team uses Purview’s quality reports for that work, the replacement needs somewhere to store and compare rule results over time. A Fabric team could use a shared results table and report. Teams running the same checks across several platforms may need a data quality or observability tool that can pull the results together.
Fabric IQ’s ontology, currently in preview, gives the business context a place in Fabric. A team can model Claim and Payment as entities and represent how a payment relates to a claim. An agent can then use those relationships when answering questions. The steward still needs a process for approving the official definition of a paid claim, relating that definition to fields across platforms, and notifying teams when it changes.
I would make OneLake Catalog the entry point for Fabric users. Where an enterprise already runs important workloads across several platforms, I would also keep an enterprise catalog that can scan those systems and maintain the business context they share. Microsoft names Atlan, Alation, Collibra, and Informatica as integration options. If a team uses two catalogs, it should decide where stewards maintain the authoritative glossary and product definitions so the claims team has one place to change what “paid” means. Fabric governance is included in Fabric capacity. The cost of maintaining enterprise coverage depends on which other jobs and systems the organization needs to support.
I’d start the transition with a data product someone already depends on. Carry its definitions, ownership, and quality rules forward, then have a consumer find it, request access, and trace a reported number to its source. Have the steward retrieve the approval and quality history. Microsoft’s Purview self-serve analytics can publish metadata about domains, products, glossary terms, subscription requests, and quality rules and results into OneLake or Azure Data Lake Storage. That gives teams a starting inventory and a way to preserve some of the evidence they’ll want during the transition. If the approval record is part of a regulatory control, I’d review that workflow with Microsoft early, as the retirement announcement recommends for highly regulated customers.
Microsoft has already been improving discovery in OneLake Catalog. At FabCon Barcelona, it showed how users could go beyond the item name to inspect tables and columns in the catalog. Every step like that makes it easier for a data owner to give people the context they need before they start working with the data.
Microsoft Ignite is next week, November 17–20, and we may hear more about OneLake Catalog there. I’ll be listening for improvements to business glossaries and lineage across platforms. I’d like the next analyst looking for paid claims to be able to find the approved definition and trace the reversal back to its source from the same place they found the model. That feels within reach.
