Almost every list of master data management tools is written for a company that is moving to a cloud data platform anyway. The list makes sense for them. It is close to useless for the other company: the one that upgraded to SQL Server 2025, discovered Master Data Services was gone, and now has a working master data process with no software under it.
That second company is who this is for. We build one of the options below, so read it with that in mind. What we can usefully add is the part the generic lists skip: which of them can actually run where your data has to stay.
Start by writing down what MDS was doing
This sounds obvious and it is the step people skip, which is why replacements get bought that cover half the job. MDS gave you five distinct things:
- A place the shared records live, on your own instance.
- Rules that refused a value before it landed.
- A screen a data steward could use without writing T-SQL.
- A version history you could query and explain.
- Subscription views that published the result to everything downstream.
Score each candidate against those five. A catalogue scores one. That is not a criticism of catalogues, which are good products. It is the reason a Purview rollout does not close this gap, and finding that out in month four is expensive.
The shortlist
| Option | Runs where | The honest note |
|---|---|---|
| Microsoft Purview + Fabric | Cloud | Strong catalogue and classification. Not a steward-editable master data store in the MDS sense. |
| Profisee | On-prem or Azure | The established MDS replacement for Microsoft estates. Quote-based, sized for the upper mid-market. |
| Semarchy | On-prem or cloud | Master data plus governance in one platform. Capable, and a genuine implementation project. |
| Informatica MDM | Cloud-first | Enterprise suite. Salesforce completed its acquisition in November 2025 and the roadmap is in motion. |
| Reltio | Cloud only | SaaS by design. Rules it out where master data cannot leave the building. |
| Build it yourself | Your SQL Server | Cheap to start. The rules engine, approval flow, history and steward screen are the part nobody estimates. |
| Primentra | Your SQL Server | Built for MDS estates. Flat published price, migration wizard, installs in about a day. |
The build-it-yourself option deserves a straight answer
A competent SQL Server team can build a master data store in a fortnight. Tables, keys, a few constraints, a stored procedure to load it. That part is genuinely small, and anybody telling you otherwise is selling something.
The part that is not small is everything a steward touches: a rules engine somebody can change without a deployment, an approval flow with a queue and a reject reason, a history that reads in English rather than as a row in an audit table, a grid that handles a four hundred row paste, and a permission model that survives an auditor. Then it needs maintaining by the person who wrote it, who will eventually leave.
Build it if master data is your product. Buy it if master data is your plumbing.
What to test, and how to make the test fail
A demo shows you a tool working. That is the least useful thing to watch. Load your own worst data and try to break it:
- Paste a block of rows containing a duplicate code. Does it refuse, and does it say which row?
- Import a file with a mandatory field blank. Does it stop, or does it write an empty string?
- Post the same bad value through the API. This is the one that catches tools whose rules live in the screen.
- Ask a colleague who has never seen the product to find who changed a supplier last month.
- Upgrade the instance. A tool that pins your SQL Server version is how you got here.
Test three is the one that separates the field. Rules enforced by a form are enforced for the people using the form, and nobody else. More on that in garbage in, garbage out.
On price
Most vendors in this category quote rather than publish, and the quote is usually per user, per domain, per record or per connector. Every one of those meters bills you more as the tool succeeds, which is a strange thing to agree to for plumbing.
Ask for the year-three number, with the domains and users you expect by then, in writing. If it cannot be produced, that is your answer about how the renewal conversation will go.
Common questions
Is Microsoft Purview a replacement for MDS?
Not for what MDS did. Purview catalogues and classifies data across an estate. MDS held the records and let a steward edit them under rules. Both are useful; only one of them is where a supplier name gets decided, and it is not Purview.
Can we keep MDS running on an older SQL Server?
For a while, on a supported version. The problem is that it pins the whole instance: you cannot upgrade the SQL Server while MDS is still on it. That turns a master data decision into a database roadmap decision, which is how these projects end up urgent.
Do we have to move to the cloud?
No, and be careful with anyone who says the question is settled. Plenty of estates cannot move master data off-premises for regulatory or contractual reasons, and the on-premises options are fewer but they are real.
What should we actually test in a trial?
Load your own worst data, not the sample set. Then break it deliberately: paste a duplicate code, import a file with a missing mandatory field, post a bad value through the API. A tool that only refuses bad data in the screen has four ways around it.
Related reading: data governance tools, honestly compared, the MDS end-of-life checklist, and master data governance.
Try the five tests on your own MDS data
Primentra reads your existing MDS models and maps the entities across, then runs on the same SQL Server. Rules apply on every route in, including the API. €7,500 per year flat, published, 60-day trial.