Nobody Told Me This About Data Governance (Until It Was Too Late)

I've been building data governance frameworks for four years. I've worked with universities, AI startups, and FMCG distributors. And every single governance initiative I've seen fail — including two of my early ones — failed for the same reason.

Not because the tools were wrong. Not because the data was too messy. Not because leadership didn't buy in.

They failed because governance was treated as a documentation project instead of a behaviour change project.

What a governance framework actually is

Most teams build a governance framework and think they're done when they have:

• A data dictionary
• A data catalogue
• Ownership assignments
• Some quality rules

That's not a governance framework. That's a file. A governance framework is the set of behaviours, processes, and incentives that make people actually use the file.

The data dictionary nobody updates is worse than no data dictionary — because it creates false confidence. Someone makes a decision based on a definition that's 18 months out of date, and you've now got a governed mistake instead of an ungoverned one.

The one thing that actually works

At MQ Centre of Applied AI, the governance framework I built didn't succeed because it was comprehensive. It succeeded because I made it unavoidable.

Every dashboard had lineage documented in the same place stakeholders went to check metrics. Every quality issue had a resolution owner and a target date in the same system the team used for sprint planning. Every definition change triggered a notification to the people using that definition in reports.

Governance that lives in a SharePoint folder nobody visits is not governance. Governance that interrupts your existing workflow and forces a decision — that's governance.

What to do if you're starting from scratch

Start with pain, not completeness. Find the one metric that causes the most arguments in leadership meetings — "revenue," "active users," whatever it is — and govern that one thing completely. Document the definition, who owns it, where it comes from, and what to do when it breaks.

Then do it again for the next most painful metric.

A governance framework with two fully trusted metrics beats a catalogue with 200 definitions nobody believes.