Data governance fails most often at the boundary between design and operations. A framework that looks complete on paper can still leave accountabilities unclear, controls unowned and lineage unevidenced.
The fix is to design governance around decisions, not documents. Start from the decisions the business and regulators need to make, work back to the data those decisions depend on, and assign ownership at that level.
Then make governance operational: embed controls in existing processes, give owners real authority, and report on data quality the way you report on financials. Governance that is measured gets managed.
Finally, sequence for credibility. Demonstrate one or two high-value use cases end-to-end before scaling. A small, working example beats a large, theoretical framework every time.