Data Core · Data Governance · Module
Definitions and lineage
What each term means, where the figure comes from, and what happens to it on the way — the record that makes two reports reconcilable.
The idea
How it works
Two reports disagreeing about the same figure is a definition problem or a lineage problem, and without a record of either it is settled by whoever argues longest.
Lineage matters as much as definition. The same definition applied to data that was filtered differently upstream produces different answers, and the difference is invisible in the report.
Working with it
In practice
- 01
Maintain one definition set
Terms defined once, referenced everywhere. Local definitions are how the argument starts.
- 02
Record the path from source to report
What is joined, filtered and transformed. This is where most discrepancies actually originate.
- 03
Make definition changes an event
With a version, a date and a note about where the trend breaks.
- 04
Name who decides a definition
One person per term, or the definition set becomes a negotiation.
One level in
The components of definitions and lineage
A component is something that exists afterwards which did not exist before — a deliverable or a mechanism, not an intention.
Most reporting discrepancies are lineage problems rather than definition problems, and lineage is rarely recorded.
The other modules in data governance
Ownership
Who is accountable for each data set being right, current and appropriately used — one person, in the business rather than in a technical function.
LearnAccess
Who may see what, on what basis, and how that is reviewed — balancing the cost of restriction against the cost of exposure deliberately.
LearnRetention and deletion
How long each kind of data is kept and what causes it to be removed — the decision that is always deferred and always accumulates.
Learn