Microsoft MDS is gone from SQL Server 2025 — what to do before October 2028

Microsoft removed Master Data Services from SQL Server 2025. Gone from the installer, no replacement, no migration tool from Microsoft. If you have built data governance processes on MDS, one version is left: SQL Server 2022, supported until October 2028.

That sounds like a lot of time. It is not. A migration starts with an inventory of everything you have built in MDS — entities, attributes, hierarchies, business rules, subscriber views, downstream integrations. That inventory alone takes weeks at most organizations, because MDS implementations grow organically over years and nobody keeps complete documentation.

What “removed” actually means

Microsoft distinguishes between deprecated features and discontinued features. Deprecated means the feature still ships but may be removed in a future release. Discontinued means it is gone now.

MDS is on the Discontinued Features list for SQL Server 2025. The SQL Server 2025 installer does not include the MDS configuration wizard, the MDS web application, or the MDS Add-in for Excel.

Microsoft has not published a replacement product or a migration guide for organizations moving off MDS. The announcement was quiet — a single entry on a technical features page, not a blog post or migration campaign. Many teams running MDS today still do not know.

The real planning window

SQL Server 2022 extended support ends in October 2028. Thirty months. That timeline compresses fast once you factor in evaluation, procurement, and actual migration work.

The organizations now beginning to evaluate alternatives will finish that evaluation in Q3 or Q4 2026. Procurement, legal review, and budget approval — realistically another quarter. That puts the start of actual migration work at early 2027. A mid-complexity MDS model takes 6 to 9 months to migrate properly: model mapping, data loading, parallel running, re-pointing downstream integrations, user training, cutover.

If your MDS implementation is large — dozens of entities, derived hierarchies, multiple systems reading subscriber views — you are looking at 12 months of migration work. Add the evaluation and procurement cycle and you need to start now, in 2026, to finish comfortably before October 2028.

There is also a less obvious cost to staying put: every month you keep MDS, you keep building on it. New entities, new attributes, new integrations. Each one makes the eventual migration more expensive. The longer you wait, the bigger the job becomes.

What you are actually choosing between

There are three realistic paths.

Most teams we talk to rule out enterprise MDM quickly. The pricing is real — Informatica and SAP MDG implementations regularly exceed €200k in the first year once you add the system integrator. The 18-to-24-month deployment timeline also means you are cutting it close against the October 2028 deadline.

The mid-market platforms are where the practical options live. Profisee has a free MDS migration utility and a well-documented migration path. Primentra was built specifically for organizations in this situation: on-premise SQL Server deployment, entity and hierarchy models that map directly from MDS concepts, and a migration wizard that handles the mechanical work of exporting your MDS model and recreating it.

Neither will do the work for you. But both give you a structured path rather than a blank page.

What migration actually involves

The same five phases apply regardless of which platform you move to.

Steps 1 and 2 — inventory and subscriber view mapping — are where most migrations take longer than planned. MDS models get extended over the years by people who are no longer at the organization. Subscriber views feed downstream systems whose integration was built by a consultant in 2018 and nobody has touched since. You find these things during the inventory, not before.

Step 4 — data validation — often surprises teams that think their MDS data is clean. MDS enforces some constraints but not all. Moving to a modern MDM platform with stricter validation rules surfaces data quality issues that MDS was silently allowing through. Those issues need to be resolved before the cutover, which means time.

Budget generously for step 5. A parallel run of 4 to 8 weeks is the minimum for a production system. The teams whose workflows depend on MDS data need time to confirm that the new platform produces the same outputs before you shut down the old one.

Where to start this week

Start with the basics, this week:

Count your entities and attributes. Open MDS and list every entity, how many attributes it has, and whether it uses derived hierarchies or explicit hierarchies. That number tells you roughly how complex your migration will be. Under 10 entities with flat structures is a straightforward job. Over 30 entities with nested hierarchies and subscriber views is a 9-to-12-month project.

Map your subscriber views. Run a query in MDS to list all configured subscriber views. For each one, identify which downstream system reads it and who owns that system. This list will drive most of your migration complexity assessment.

Start a vendor shortlist. Do not wait for a formal RFP process. Get trial access to two or three platforms, import a sample of your MDS data, and see how the entity and hierarchy models map. The differences become clear quickly when you work with real data rather than feature comparison matrices.

Frequently asked questions