Data governance is about ownership, not committees
Why most data governance initiatives stall, and how to move from a formal committee to real accountability.
Data governance is one of those concepts that companies periodically "implement" without actually changing anything. A committee is formed, a policy is drafted, a document is approved. A year later nobody remembers when the committee last met, and the problems with data quality and availability are exactly the same as before.
I have seen this cycle often enough to articulate why it keeps repeating - and why real data governance works differently.
What is wrong with the committee approach
Data governance committees are usually created with the right intentions. Representatives from business units, IT, and sometimes legal gather. They agree on rules. They assign responsibilities.
The problem is that accountability in this model is collective. And collective accountability in practice means nobody is specifically accountable. When data turns out to be poor quality, everyone looks at the committee. When the committee meets once a quarter, many problems accumulate in the meantime that nobody resolved.
On top of that, a committee typically deals with policies and standards, not operational reality. It describes how things should be but has no mechanism to make that happen every day.
What real data ownership looks like
Real data governance starts with a simple question: who is the specific person responsible for these data being correct, current, and available?
Not a role. Not a department. A specific named person who knows that if the data breaks, they will be the first one asked about it.
In mature teams this is called a data owner or a steward, depending on the level of responsibility. A data owner is typically someone at the director level who makes business decisions about how data is collected and used. A data steward is the person who handles the operational side: monitors quality and answers questions from consumers.
Without these roles filled with real names, any policy remains paper.
Three things without which data governance does not work
First - explicit owners for each critical data set. Not just a name in a document, but people who know they are owners and understand what that means.
Second - processes for responding to data problems. Where does a ticket go when an analyst finds a discrepancy? Who makes the correction decision? How is it recorded?
Third - quality metrics that someone reviews regularly. Not once a quarter at the committee, but weekly or daily depending on criticality. Without measurement there is no management.
How to start without a big project
The right first step is not to create a committee or write a policy. The right first step is to take two or three of your most critical data sets and for each one answer these questions:
- Who right now is de facto watching the quality of this data?
- What happens when it breaks - who notices and who fixes it?
- Is there documentation for the structure and semantics of this data?
- How do consumers of this data learn about changes?
If there are answers to these questions, you already have the seed of real data governance. The task is to make it explicit and extend it to the other critical data sets.
If there are no answers, that is exactly where to start - not with writing a policy.