The two phrases get swapped for each other in almost every meeting where they come up. It stays harmless right up until a budget is approved, at which point somebody buys the layer they did not need and the other problem is still sitting there six months later.
I have watched this happen twice, both times to people who were not confused about their own data. They were confused about which of the two words the vendor in front of them was using. The distinction itself is not subtle. It gets blurred by companies who sell both halves and would rather you did not separate them.
The short version
Data governance decides. Who owns the supplier domain. What a valid VAT number looks like. Who signs off a bank detail change. How long the history is kept. It covers every kind of data, not only the shared records, and most of it is agreements rather than software.
Master data management stores and enforces. It is where the supplier record lives, where the VAT rule refuses a bad value, where the bank detail change waits for its approver, and where the history is written. It covers the shared records only, and almost all of it is software.
| Data governance | Master data management | |
|---|---|---|
| What it is | A set of decisions and owners | A system holding shared records |
| Covers | All data in the organisation | Customers, suppliers, products and the like |
| Mostly made of | Agreements, roles, policy | Software |
| You know it works when | Somebody can say who decides | A bad value gets refused |
| Fails as | A document nothing enforces | A tidy home for unagreed values |
| Bought from | Consultants, or written in-house | A vendor |
Governance without a system
This is the common half, because it is the half you can start on Monday without a purchase order. A framework gets written. A stewardship committee meets. A quality dashboard appears with a number on it.
Nothing refuses anything. The rules are correct and live in a PDF, and the overnight load has never read a PDF. Two years later the dashboard number is roughly where it started, and the programme gets quietly folded into someone's objectives.
The tell: ask what the last thing was that the governance programme prevented. Not measured, not reported. Prevented.
A system without governance
Rarer, and it fails in the opposite direction. The tool goes in, the records load, the screens work, and the arguments carry on exactly as before, because nobody named an owner and nobody wrote down what a valid record looks like.
The tell: the tool is in production and two departments still disagree about which supplier name is right. Storage was never the disagreement.
Do a thin slice of both
The usual advice is that governance comes first. It is half right, and the half that is wrong costs a year. Deciding an owner and five rules for one domain takes days, not quarters. Put those five rules into something that enforces them and you have both layers on one domain inside a month.
Then the second domain is an easier conversation, because you can point at a count of refused values rather than a maturity model. There is more on how to run that first slice in master data governance.
Common questions
Our consultants say governance comes first. Are they wrong?
Half right. Deciding an owner and a handful of rules has to come first, and that takes days. Spending a year producing a framework before anything enforces anything is how the budget runs out before the first bad value gets refused.
Is a data catalogue governance?
It is a tool that supports governance and does not perform it. A catalogue describes what exists. Governance decides what is allowed. A perfectly catalogued estate can still have four addresses for one supplier.
Where does data quality sit?
It is the measurement. Governance sets the standard, master data management enforces it, quality reports how far off you are. Buying only the measurement is the most common way to spend three years learning your data is bad.
Do we need both if we are small?
You need an owner and a few rules, which is governance whether or not anyone calls it that. Whether you need a system depends on one thing: how many places the same record is created in. One place, a spreadsheet may hold. Three places, it will not.
Related reading: data governance tools, honestly compared, and MDM vs data catalog.
The bottom layer, on your own server
Primentra is the enforcement half: records, rules that run on every route in, approval on the fields that matter, and a history a non-technical colleague can read. Your decisions stay yours. €7,500 per year flat, 60-day trial.