
Microsoft Purview - The (Broken) Promise of Unified Data Governance
Last week I was having lunch with a contact at Microsoft and they mentioned that Purview (specifically the structured data governance product) was due to be retired and replaced by OneLake Catalog in Fabric. Later, I saw that Microsoft had made the official announcement.
I must have let out the proverbial sigh and thought "I've seen this film before". I have gone through the same process with products like ProClarity, PerformancePoint, Azure Data Catalog, and Master Data Services. It would be unfair to say that only Microsoft does this; vendors like Google, Qlik, and Salesforce are guilty of this too - getting customers to go all in on a new technology and then leaving them high and dry.
We all know that the intrinsic value to organisations is not in the technology itself, but in the knowledge and capability that is established in the organisation through both the technology and organisational change efforts.
The implication of this is significant. Between 2011 and 2015 I worked on at least five large data transformation projects, all relying on PerformancePoint and Master Data Services. One of these programmes had invested £1.4m, meaning that yet another large investment was required to replace these products, sooner than the organisation had planned to.
So what does this mean for organisations already using Purview and for organisations that have it on their 2027 data management roadmap?
In this article, I am going to break down:
- The exact changes that are coming
- What this means for organisations already using Purview
- What this means for organisations planning on using Purview
- How your data strategy can mitigate against the next technology that is sunsetted
What Has Actually Been Announced
| Microsoft Purview Feature | What Happens |
|---|---|
| Unified Catalog which is the glossary, data products, and governance domains. | These will all be retired but available in maintenance mode until 12 November 2027. |
| Data quality rules and scans. | These will all be retired and there was no announcement about a migration path for data quality rules. |
| Data Map scanning and lineage. | Retiring. |
| Classic Azure Purview governance. | Retiring. |
| Information Protection, DLP, DSPM, insider risk. | This stays in Purview. |
| OneLake Catalog in Fabric | This is named as the successor with the data governance features being included in Fabric capacity. |
There are a few points to make about the above:
- Microsoft's announcement says OneLake Catalog can govern data outside of Fabric through shortcuts, mirroring, and connectors. It names Atlan, Alation, Collibra, and Informatica for organisations that need more. In practice, the data has to be ingested into OneLake first, which is a different model from Purview, where sources were scanned without ingesting the data (e.g. Oracle RDBMS, Snowflake, and Azure SQL).
- Data quality is the biggest gap as the announcement doesn't mention it at all. Ironically, on the Purview projects I worked on in 2022 and 2023, this was a big gap that Purview couldn't fill. The feature finally arrived, and now it's being retired before many organisations have had the opportunity to use it.
- Microsoft has given 12 months' notice and is pointing regulated customers to their account teams. That is better than some of the technology retirements I've seen before.
What Does This Mean If Purview Is On Your 2027 Roadmap?
Well, effectively the decision has been made for you. New accounts can't be created. So the most useful question to revisit is "What were you aiming to achieve with Purview?" Go back to the intended outcomes (trusted definitions, clear ownership, measurable data quality, discoverable data) and re-scope your data management initiative from there. It is also worth considering if what you want to achieve would work as well without a tool as with one.
Next, revisit the business case. The Purview business case that you had built probably assumed adoption growing over three years, with payback achieved around year one. Ask whether any alternative tool you're now considering will be the same, better, or worse.
If you aren't planning to use Fabric, the options are an independent catalogue, the native catalogue of the platform you do use (e.g. Databricks Unity Catalog, Snowflake Horizon), or a deliberately lightweight approach for a year while the market settles. Review the pros and cons of each technology option (including none at all) and also think about the likelihood of one of the vendors getting acquired by another. For example, Informatica was acquired by Salesforce.
Regardless of your technology choice, use this time well. Continue to focus on the operating model, the business glossary, and business ownership as these can all progress independently of any tool.
What Does This Mean If You Are Already Using Purview?
Although 12 months seems like a long time, it will pass quickly. Spend time on the following:
- Take stock of what you've already built, and be honest about what is being actively used. When I recently reviewed an organisation's Purview estate, it had a 400-term business glossary, around 20 stewarded domains and access-request workflows, but data products, data quality and profiling had never been used, and catalogue usage was falling. For many organisations, the thing worth saving is the business glossary and the stewardship model, both of which are portable.
- Export as much as you can now, while the APIs and tooling are fully supported, and store it somewhere you control.
- Separate the decision about the catalogue from the decision about Fabric. If Fabric is already your direction, OneLake Catalog may be the simplest path. If it isn't, don't adopt a data platform technology just to keep a catalogue.
- Plan the data quality rebuild, because it has no announced migration path.
- Include a business change workstream. Data stewards who spent time learning Purview, for example, will need to go through the change process again.
- Validate what remains available: Information Protection, DLP and insider risk are unaffected, and many organisations use Purview for this reason.
How To Plan a Data Strategy That Survives the Next Product Retirement
Over the years, responding to RFPs and helping organisations understand their data management requirements, I have often heard the term "future-proof". This is a misnomer because it is impossible to be future-proof. At best, you can mitigate risk. There are several ways this can be done when defining or changing your data strategy.
- Ensure that your organisation always owns its content. The metadata, glossary, ownership model, data quality rules, and policies are organisational assets and should remain portable. Keep a copy of all of these assets in a non-proprietary format.
- Get the operating model right first. Roles, data ownership, data stewardship, decision rights, and governance forums can withstand a change of data governance tooling.
- Try to understand where a candidate data governance product sits in the vendor's overall strategy. Whilst a vendor may be loath to share their roadmap early in the product lifecycle (Microsoft certainly was), make visibility of this a non-negotiable in your procurement process. There are also several warning signs such as repeated product renames, the product being described as "integrated with" a newer platform, documentation and roadmap updates slowing down, and the vendor's own industry events no longer featuring it.
- Plan for a product deprecation event as part of the strategy. Include an exit cost and exit criteria in the business case, and negotiate notice periods, data export, and transition support in the contract.
- Match your investment timeframes to the likely lifespan of a platform. Assume a governance tool has a three-to-five-year life, and be cautious about deep customisation that ties your processes to one product's way of doing things.
- Treat people as a continuity risk too. People move around the organisation and also leave the organisation. Ensure that you have some kind of succession plan for data owners and data stewards, such as named deputies.
- Keep a technology risk register and review it every year. Each strategic tool must have a named owner who can keep tabs on the product and vendor.
- Focus on what you can control. Whilst you can't control a vendor's product strategy, ensure you know how dependant you are on the use of that product.
Ultimately, Microsoft, and other vendors, will continue to retire products. Organisations that own their metadata and related content can hope to experience these sorts of events as an inconvenience rather than a restart from zero.
As a practical next step, spend an hour or two this month listing every tool your data governance depends on and what you'd lose if that tool disappeared. Most importantly, make sure that each tool has an owner.

Günter Richter
Founder & Principal Consultant, Umlaut Consulting
30+ years of experience in strategic consulting and data transformation. Helping organisations unlock the real value of their data.
Related Posts

Why a Business Data Domain Model Is Key To Your Data Management Programme
After graduating from university, I started my career implementing ERP solutions. Whilst not the most glamorous of jobs, it was incredibly valuable because it...

Data Governance Is Not About Governance
You’ve just been made responsible for improving data governance in your organisation. So you do what any sensible person would do. You research frameworks,...

Yet Another Data Breach
At an event yesterday evening, I talked to several business owners and business leaders, from small and mid-size organisations, about their data challenges....