ADE Microsoft Purview
A growing company soon owns thousands of tables and files across many systems. Nobody remembers where the customer data sits, who owns it, or which reports depend on it. Microsoft Purview solves this problem. The service scans data sources, builds a searchable catalog, and tracks how data moves from origin to report.
The Library Analogy
A huge library without a catalog forces visitors to wander the shelves. A catalog lists every book, its subject, its author, and its shelf location. Purview acts as the catalog for company data. A search for "customer email" lands on the exact tables that hold it.
Core Capabilities
- Data Map: A live inventory of data sources and the assets inside them.
- Scanning: Automated reading of sources such as Azure Data Lake Storage, Azure SQL, Synapse, Power BI, and Fabric to collect technical details.
- Classification: Automatic tags for sensitive content, such as phone numbers, national IDs, or card numbers.
- Business glossary: Agreed definitions of terms like "active customer", linked to real tables.
- Lineage: A picture of how data flows from sources through pipelines to reports.
- Sensitivity labels: Protection tags that follow data into other tools.
Microsoft groups the governance experience in a Unified Catalog, where data owners publish and manage data products. Menu names change as the product grows, so check the current portal layout when you start.
How Purview Fits the Platform
DATA SOURCES PURVIEW PEOPLE
Data Lake Storage --scan--> +-------------------+
Azure SQL Database --scan--> | Data Map | ---> Search by keyword
Synapse / Fabric --scan--> | Classifications | ---> Request access
Power BI --scan--> | Glossary | ---> Check lineage
ADF and Databricks --lineage-> | Lineage | ---> Review sensitive data
+-------------------+
Scenario: Harbor Bank Audit Request
Harbor Bank, a fictional lender, receives a request to list every table that stores customer phone numbers. Before Purview, the team spent two weeks emailing data owners. After a scan, the team runs one search for the "Phone Number" classification and receives a list of tables, owners, and locations in minutes.
Scanning Step by Step
- Register a data source, such as a storage account, in the Data Map.
- Grant the Purview identity read access to that source.
- Choose a scan rule set that lists the classifications to look for.
- Run the scan once, then schedule it weekly or monthly.
- Review the discovered assets and assign an owner to each important one.
Reading a Lineage Graph
Lineage answers the question "where did this number come from?" A report shows a sudden drop in sales. The lineage graph traces the report back through a gold table, a silver job, and a Data Factory copy activity, and it pinpoints the pipeline step that changed.
Orders DB ---> ADF Copy ---> Bronze table ---> Spark clean ---> Silver table
|
Spark aggregate
|
v
Gold table ---> Power BI report
Impact analysis uses the same graph in the other direction. Before changing a column, a data engineer checks every downstream table and report that the change touches.
Roles and Responsibilities
| Role | Job |
|---|---|
| Data owner | Decides who may use a dataset and for what purpose |
| Data steward | Maintains descriptions, glossary links, and quality rules |
| Data engineer | Registers sources, runs scans, and wires pipelines for lineage |
| Data consumer | Searches the catalog and requests access |
Good Habits for Governance
- Start with the most sensitive and most used sources instead of scanning everything on day one.
- Write plain-language descriptions, since search quality depends on them.
- Assign a named owner to every important asset.
- Review classifications for false positives after the first scan.
Key Points
- Purview builds a searchable inventory of data across many systems.
- Scans classify sensitive content automatically.
- Lineage shows the journey of data from source to report.
- Owners and a shared glossary turn the catalog into a living guide.