Data Governance Program
A governance program designed to look like a product — onboarding flow, named owners, a steward UX that the people doing the work didn't resent.
- Role
- Program Director
- Year
- 2023
- Stack
- Snowflake · dbt · Looker
Outcomes
- Took ownership of 60+ orphaned data sources.
- Defined 14 stewardship contracts across faculties.
- Replaced quarterly compliance scramble with a continuous process.
Why governance usually fails
Most governance programs are designed for the auditor. They produce artifacts the auditor needs and create work that everyone else resents. The stewards become unpaid compliance clerks. Within eighteen months the program is a museum of dead policies.
A different default
We designed the program with the steward as the primary user. The auditor became a downstream consumer. That single inversion changed which surfaces got polish and which got tolerated. Stewards got a real UI. Auditors got an export.
What it looked like 18 months later
The thing that surprised me was how little of the program turned out to be about data. It was about ownership: who is allowed to make a decision about this thing, and how do we make it cheap for them to say no.