The three pieces that matter
When we work with internal research teams, the governance items that consistently deliver value are small and unglamorous. They are:
1. A data dictionary
Every field in every internal table has a one-line definition: source, unit, frequency, timezone, and the convention used. New analysts should be able to read the dictionary and understand a table without asking the person who built it.
2. Named ownership
Every dataset has one named owner, even if that owner is not the person who updates it. The owner is the person who answers "is this number still right?" When ownership is diffuse, datasets quietly drift out of date and no one notices.
3. A known-issues register
A short list of the things we know are wrong or approximate in the data, and why. This sounds small, but it is the difference between a dataset an analyst trusts and one they do not. If the known gaps are documented, the analyst can decide whether the gap matters for their specific question. If they are not documented, the analyst must assume there may be gaps everywhere.
We have seen these three items unblock more research work than any new visualisation or dataset. The dashboard can wait; the dictionary cannot.