"Which data governance tool should we buy?" is one of the most common questions I get, and one of the most consistently mis-framed. People arrive expecting a feature bake-off — a spreadsheet of capabilities where the tool with the most ticks wins. I almost never answer it that way, because after enough of these decisions you learn that the governance tool you can actually operate beats the one with the best feature matrix every single time. The right catalogue is mostly a function of where your data already lives, who's going to run the thing day to day, and whether you'll feed it — not which vendor's demo dazzled hardest. So let me give you the honest decision framework instead of a scorecard.

The three options I see most are Microsoft Purview, Databricks Unity Catalog, and a dedicated third-party catalogue like Collibra or Alation. They're not really competing for the same job, which is the first thing to understand.

Microsoft Purview: the default when you live in Microsoft

If your data estate is centred on Azure and Fabric, Purview is the path of least resistance, and least resistance in governance is worth more than it sounds — because the governance tool that's already wired into your platform is the one people might actually use. Purview's tightening integration with Fabric means lineage, sensitivity labels, and cataloguing that flow with the data rather than sitting in a separate system someone has to remember to update. When I called Purview's GA a new era of business-led governance, that integration was the reason. For a Microsoft-centric organisation, the burden of proof is on not choosing it.

The honest caveats: Purview is at its best inside the Microsoft world and less compelling as a truly heterogeneous, multi-cloud catalogue spanning lots of non-Microsoft sources. And "integrated" is not the same as "automatic" — it still needs owners, curation, and a steward who cares. An unfed catalogue is a stale catalogue whatever the badge on it.

Unity Catalog: the default when Databricks is your centre

If your centre of gravity is Databricks — a lakehouse-first, data-science-heavy shop — then Unity Catalog is the natural governance layer, because it governs at the place the data and compute live. Governance that's native to your primary platform is enforced governance; governance bolted on beside it is aspirational. For a Databricks-first organisation, reaching past Unity Catalog to a separate tool usually adds seams without adding control.

The caveat mirrors Purview's: it's strongest within its own world. If Databricks is one platform among several rather than your core, Unity Catalog governs its slice well and the rest not at all.

A dedicated catalogue: when neutrality is the requirement

Then there's the class of organisation for whom neither platform-native option fits — genuinely heterogeneous estates, many clouds and tools by deliberate strategy, or a regulatory posture that wants the catalogue independent of any one platform vendor. Here a dedicated third-party catalogue earns its keep: it's built to sit above everything and be nobody's captive. The trade is real, though. You're taking on a separate product to license, integrate, staff, and keep fed — the integration tax I keep warning about — in exchange for breadth and independence. For a large, truly multi-platform enterprise with the maturity to run it, that trade is often right. For a mostly-Microsoft mid-size organisation, it's usually buying a Formula 1 car to do the school run.

How I actually decide

The questions I ask, in order:

  1. Where does your data actually live? Not aspirationally — actually. The catalogue should sit where the centre of gravity is. Microsoft-centric → Purview. Databricks-centric → Unity Catalog. Genuinely everywhere → consider dedicated.
  2. Who will operate it? A catalogue is a product with owners and stewards, not a switch. If you can't name the people who'll curate it, the fanciest tool will rot. Choose the one your team will realistically feed.
  3. How heterogeneous are you, honestly? Most organisations that call themselves "multi-cloud" have one dominant platform and a long tail. Govern the dominant platform natively and reach for neutrality only if the tail is genuinely load-bearing.
  4. What does the regulation demand? If you have an obligation that specifically wants independence or cross-platform lineage — some high-risk AI and heavily regulated contexts do — that can override the convenience of the native option.

The recommendation, plainly

For the majority of the organisations I advise — Microsoft-centric, running Fabric or Azure, mid-to-large, without a full-time governance function to spare — I recommend Purview, and I recommend it not because it wins a feature war but because it's the one that's already there, already integrated, and therefore the one that might actually get used. The best governance tool is the one your organisation operates, and integration is the single biggest predictor of whether it will. For a Databricks-first shop the same logic points at Unity Catalog. And for the genuinely heterogeneous enterprise with the maturity to run it, a dedicated catalogue is a defensible, sometimes necessary, choice.

What matters more than the tool choice, though: no catalogue governs anything by itself. I have watched organisations buy the most capable governance platform on the market and end up with a beautifully-licensed, comprehensively-empty catalogue, because they treated the purchase as the finish line rather than the starting gun. The tool is maybe twenty per cent of the outcome. The other eighty is ownership, stewardship, and a culture that keeps it fed — and that part is identical whichever logo you choose. So by all means agonise over Purview versus the alternatives. Just don't mistake it for the decision that actually determines whether your data gets governed. That decision is about people, and no vendor can sell it to you.