Two Lenses on Purview in MS Fabric

Microsoft uses one brand—Purview—for two scopes. After FabCon Europe (Vienna, Sep 15–18, 2025), the split is even clearer:

  • Enterprise Purview (in the Purview portal) is your estate‑wide governance, security, and compliance plane: unified catalog, lineage, data quality, sensitivity labels/DLP/audit, DSPM for AI, and (now) Fabric‑aware risk insights.
  • Purview inside Fabric (OneLake catalog + Purview hub) is the in‑product lens for people building and consuming inside Fabric. The Govern tab is now GA, with domain‑scoped insights; domains have public APIs.

Think of it this way: Fabric’s lens governs work where it’s made; Enterprise Purview governs the whole estate.

Catalog & discovery

The enterprise problem isn’t “find my dataset in this workspace”; it’s “make the whole estate findable and understandable.”

  • Enterprise Purview (Unified Catalog). The catalog sits on the Data Map and—post‑FabCon—adds two practical boosts for Fabric estates:
    (a) Sub‑item metadata for Fabric Lakehouses (preview): table/column/file‑level metadata now appears in Purview for richer discovery.
    (b) Custom business attributes (preview): catalogue assets in the language of your domains/products/glossary.
    These features tighten discovery without forcing everyone into Fabric first.
  • Inside Fabric (OneLake catalog). The Explore view keeps makers fast; the Govern tab (now GA) surfaces posture and recommended actions in context, and you can scope insights by domain. Use it for day‑to‑day ownership inside Fabric; jump to Purview when the question crosses platforms.

What changed at FabCon Europe? Granular metadata for Lakehouses helps people find and understand Fabric data from the enterprise catalog, and Fabric’s Govern tab went GA, making local governance work less “portal‑hopping.” (Remember: sub‑item metadata ≠ sub‑item lineage; see lineage section.)

Lineage

The enterprise problem is blast‑radius analysis across systems: “If I change this source, what breaks downstream” along with “Where did this data come from, can I trust it?”

  • Enterprise Purview (estate‑wide lineage). Purview stitches relationships across sources and services and can ingest Fabric items once you register/scan your Fabric tenant. It’s the right lens when lineage spans multiple workspaces, services, or clouds.
  • Inside Fabric (workspace lineage). Fabric’s lineage view is intentionally workspace‑centric—great for local impact checks by creators. It’s not trying to be the cross‑cloud graph. (That’s the enterprise lens.)

Expectation vs. reality after FabCon.
FabCon’s sub‑item metadata for Lakehouses improves discovery depth, but Microsoft’s docs still note: for most non–Power BI Fabric items, Purview scanning collects item‑level lineage; Lakehouse sub‑item lineage is not supported yet. In other words, you can now see columns in the catalog, but you generally can’t follow columns end‑to‑end. Calibrate your audit stories accordingly.

Data quality

Quality has two audiences:

  • Where engineers work (Fabric). Fabric gives you profiling/observability in the product. New from FabCon: published error records in Fabric OneLake (preview) so owners can investigate and remediate in place, closer to where errors occur.
  • Where stewards govern (Enterprise Purview). Purview’s Data Quality evaluates rules and rolls up scores at asset/product/domain levels—so leaders can prioritize and track improvement across the estate (including Fabric assets you scan into Purview).

Why this matters. The Fabric‑side “fix it here” experience and the Purview‑side “measure it across domains” experience finally meet: errors show up where makers live, while quality scores live where stewards decide.

Compliance & protection

Compliance must be uniform; analytics tools should honor it—not reinvent it.

  • Enterprise Purview (control plane).
    GA at FabCon:
    • Information Protection policies for Fabric items—manual labeling with enforcement anchored in central policy.
    • DLP for structured data in OneLake—now GA across Fabric (not just semantic models).
    • Insider Risk Management indicators for Power BI—(view/download/export/label actions as risk signals).
    • Security & compliance controls for Copilot in Power BI—discover/triage sensitive content in prompts/responses with governance hooks (audit, eDiscovery, retention).
    Preview: Data Risk Assessments for Fabric in DSPM for AI—auto‑flag oversharing (default assessment targets your top 100 accessed Fabric workspaces).
  • Inside Fabric. Fabric applies those controls: items carry sensitivity labels, DLP evaluates supported items, and activity flows to the unified audit in Purview. Microsoft’s “What’s new” confirms DLP for Fabric and Power BI is GA this month.

Net effect. The compliance brain stays in Purview; Fabric participates by honoring labels, enforcing DLP, emitting audit, and—now—being assessed for oversharing risk via DSPM for AI.

Putting the lenses to work (post‑announcements)

  • Discovery & shared language → Enterprise Purview Unified Catalog. Use domains, data products, glossary—and (now) custom attributes—to describe Fabric and non‑Fabric data in business terms. Use Fabric’s Govern tab (GA) for your items’ posture; switch to Purview to see and curate the estate.
  • Impact analysis → Two scopes of lineage. Fabric lineage is for workspace impact; Purview is for estate impact. The FabCon update gives you sub‑item metadata to enrich search, but lineage remains largely item‑level for Lakehouse. Don’t promise column‑level traceability where it doesn’t exist.
  • Trustworthiness → Quality in two places. Use Fabric’s “published error records” (preview) to fix data where it lives; use Purview Data Quality to measure and improve across domains and products.
  • Compliance → Centralized definition, ubiquitous enforcement. Define labels/DLP/Audit/DSPM in Purview; expect Fabric to enforce and surface signals. With DLP now GA across Fabric and risk assessments lighting up for Fabric workspaces, the end‑to‑end story is more complete.

Bottom line for readers

Use Fabric’s Purview features to keep governance close to creation and consumption. Use Enterprise Purview to govern the estate: a unified catalog (now with deeper Fabric metadata and custom attributes), cross‑platform lineage, stewardship‑grade data quality, and compliance that’s consistent everywhere (now with GA DLP for Fabric, Insider Risk for Power BI, and Fabric‑aware DSPM). On lineage specifically, success still hinges on right‑sizing expectations: rich sub‑item discovery ≠ sub‑item lineage—for that, Purview remains primarily item‑level with Lakehouse today.

Unknown's avatar

Author: Jason Miles

A solution-focused developer, engineer, and data specialist focusing on diverse industries. He has led data products and citizen data initiatives for almost twenty years and is an expert in enabling organizations to turn data into insight, and then into action. He holds MS in Analytics from Texas A&M, DAMA CDMP Master, and INFORMS CAP-Expert credentials.