Unity Catalog vs Purview: Pricing, Lineage and Which to Buy for Databricks

Unity Catalog has no price of its own. Microsoft Purview does. Unity Catalog is included with Databricks Premium, so its real cost is the Premium tier uplift on compute: Jobs Compute goes from $0.15 to $0.30 per DBU-hour, exactly double. Purview bills separately at $0.0165 per governed catalog asset per day plus $15.00 to $240.00 per Data Governance Processing Unit. Both figures come from the public Azure retail price API, read on 30 August 2026.

That single difference in shape, a bundled tier uplift versus an itemized per-asset meter, drives most of what follows. One scales with how much compute you run. The other scales with how many things you catalog. Here is how they compare on price, on lineage coverage, and on the estates each one actually suits.

The short answer

What Unity Catalog actually costs

Search the Azure retail price API for a Unity Catalog meter and you will not find one. There is no SKU, no per-asset rate, no line item. Unity Catalog is a Premium tier feature, which means its price is hidden inside the difference between Standard and Premium DBU rates.

Read live on 30 August 2026 for Azure Databricks in East US:

All rates per DBU-hour, US dollars, region East US, read from the public Azure retail price API on 30 August 2026. List prices before commitment discounts.

Three things fall out of that table. The uplift is nowhere near uniform: Jobs Light Compute costs 3.1 times more on Premium while SQL Analytics costs exactly the same on both tiers. So a team whose workload is mostly SQL warehouses pays almost nothing for governance, and a team running lots of light batch jobs pays the most. Serverless SQL at $0.70 per DBU-hour is Premium-only, so there is no Standard comparison to make.

To be fair to Databricks, Premium buys more than Unity Catalog. It also includes role-based access control, audit logs, credential passthrough and conditional access. Attributing the whole uplift to governance overstates it. But the direction is right, and it is the number you should model, because you cannot buy Unity Catalog without buying Premium.

A worked example. A team burning 20,000 Jobs Compute DBU-hours a year pays $3,000 on Standard and $6,000 on Premium. The governance decision costs about $3,000 a year. That is genuinely cheap, and it is the strongest argument for staying native if your estate is Databricks-only.

What Microsoft Purview actually costs

Purview goes the other way and itemizes everything. The Data Governance family carries four meters, unchanged since 1 November 2024 and identical in all 23 regions we checked:

Read from the public Azure retail price API on 30 August 2026, region East US.

The asset meter is the one to model carefully, because the daily rate disguises the annual number. At $0.0165 a day, one governed asset costs $6.02 a year. Ten thousand assets cost $60,225 a year before you provision a single processing unit. Thirty thousand assets cost $180,675, which is more than the $170,000 Collibra lists on AWS Marketplace for its platform.

Note also that Basic, Standard and Advanced processing units are all sold as one Data Governance Processing Unit in the same unit of measure, with a 16x spread between the cheapest and the dearest. Which tier your scans consume moves a Purview bill more than almost anything else. The full meter set, including Data Security and Data Compliance, is in our breakdown of Microsoft Purview pricing.

Which one is cheaper?

On a Databricks-only estate, Unity Catalog wins and it is not close. Governance costs a few thousand dollars a year as a compute uplift against tens of thousands as a per-asset meter, and Unity Catalog is already switched on.

The comparison inverts as soon as your obligations extend past the platform. Purview is priced for an estate, not a warehouse. If you also have to label documents in SharePoint, run eDiscovery on mailboxes, apply retention policies and prove all of it to an auditor, no amount of Unity Catalog covers it, and buying separate tools for each of those jobs costs far more than the Purview meters. The per-asset rate stops looking expensive when you count what it replaces.

The trap is the middle case: a team that buys Purview purely to catalog a Databricks estate that Unity Catalog already catalogs for free. That happens more often than it should, usually because governance was scoped centrally without anyone checking what the platform already does.

Lineage: what each one actually captures

Both draw lineage graphs, and both have published limits worth knowing before you commit.

Unity Catalog captures lineage automatically for anything that runs on Databricks, down to the column level, across every workspace on the metastore, covering tables, views, notebooks, jobs, pipelines, queries, dashboards and ML model versions. Databricks documents the gaps honestly: column lineage does not resolve for path-based reads such as select * from delta."s3://bucket/path", user-defined functions blur the column-to-column mapping, and lineage is not preserved when you rename a catalog, schema, table, view or column. The system tables system.access.table_lineage and system.access.column_lineage keep a rolling one-year window, while Catalog Explorer retains lineage indefinitely for data captured after 1 September 2024. We go deeper on all of it in Unity Catalog lineage.

Purview publishes a per-source capability matrix, and it is the most useful page in its documentation. Amazon Redshift is marked No for lineage, and for classification, labels, policies and live view as well. Snowflake lineage is static, generated from object definitions among tables, views, streams and stored procedures rather than from executed queries, incremental scans are not supported so every scan is a full scan, and stored procedure lineage does not cover procedures written in Java, Python or Scala. For Power BI, Microsoft states that column level lineage and transformations other than Dataflows are supported when using Azure SQL Database as the source, and that other sources are not currently supported. The details are on our Microsoft Purview comparison.

The pattern is clear enough. Unity Catalog is deep and cheap inside Databricks and blind outside it. Purview is broad across Microsoft and thin on non-Microsoft warehouses, especially Redshift.

Do I need both Unity Catalog and Purview?

If you have Microsoft 365 compliance obligations, yes, and they layer rather than overlap. Purview scans Azure Databricks Unity Catalog as a registered source and pulls its metadata into the wider tenant catalog, so the two are designed to work together. Unity Catalog governs the platform, Purview governs the organization.

If your obligations stop at the analytics platform, no. Buying Purview to catalog a Databricks estate is paying a per-asset meter for an inventory you already have.

There is a third case that neither handles well: a regulated team that has to answer questions about individual people rather than about tables. A subject access request is not a catalog query, it is a hunt for every system holding one person's records, and dedicated software for locating where a person's data lives solves that far more directly than a metadata catalog does. Catalogs tell you what a table is. They do not tell you whose data is in row 40,000.

What neither one does

Both tools are backward-looking. They tell you what happened and where data came from. Neither answers the question an engineer asks on a Tuesday afternoon: if I drop this column tomorrow, what breaks?

That is a different computation. It needs a lineage graph built from executed queries rather than declared object definitions, so it follows ad hoc CREATE TABLE AS SELECT statements, dbt models and backfills the same way, and it needs to be run against a proposed change rather than a past one. Datatrail connects read-only to Snowflake, BigQuery, Databricks or Redshift, parses query history into column-level lineage, and turns it into impact analysis: every downstream model, exposure and dashboard that reads the field you are about to change, listed by name, before you merge.

It runs alongside both. Keep Unity Catalog for platform governance, keep Purview for tenant compliance, and add lineage that spans warehouses when your estate outgrows a single platform. The wider landscape is in our data lineage tools and data governance tools comparisons, and if you are weighing the native option against a dedicated one, Databricks Unity Catalog compared covers that trade directly.