
Many governance programmes fail because they start with committees and policy documents. People route around them, and the data stays just as unreliable.
Begin with ownership
Every important dataset needs a named owner who can answer three questions: what does it mean, who may use it, and is it right? Without that, no tool will help.
Agree on definitions that matter
Pick the dozen terms that cause the most arguments, such as “active customer” or “net revenue”. Write one definition for each, store it where analysts will see it, and use it in the shared models.
Automate quality checks
- Test for freshness, completeness and duplicates inside the pipeline.
- Alert the owner when a check fails, not a distant central team.
- Show the status next to the dataset so users know what they can rely on.
Control access with the least friction
Classify sensitive data once, then apply access rules by role. People should be able to request and receive access in a predictable way, with a record kept for audit.
Add process only when it solves a problem
Each new rule should answer a real pain: a regulatory duty, a repeated error, a security risk. If nobody can name the problem, the rule is probably bureaucracy.
A first month
- List the most-used datasets and assign owners.
- Write down the dozen contested definitions.
- Add three automated quality checks to the busiest pipeline.
Want to put this into practice?
Talk to our team about your data, planning and AI goals, or book a demo and see how it works on your own data.


