Governed Metrics and Definitions
Governed Metrics and Definitions
This domain tests metric governance: one agreed definition, certified for reuse, with explicit dimensionality and filters. Questions often ask why two reports show different values for the same business term.
Core rules
- A governed metric has an owner, a definition, a formula, a grain, and a certification status. Reuse it rather than recreating it.
- Compound metrics reference other metrics; derived metrics combine metrics and constants; level metrics force dimensionality; conditional metrics embed a filter; transformation metrics apply offset logic.
- Dimensionality: a metric can be absolute (ignore report attributes), dimensional (follow the report), or fixed to a level.
- Ratios must aggregate numerators and denominators before dividing; average-of-averages is usually wrong.
- Report filters and metric filters are different: metric filters live inside the definition and change the metric's meaning.
Worked example
Two reports show Gross Margin %: one 42%, one 38%. Report A uses Avg(Margin %) at row level; Report B uses Sum(Profit)/Sum(Revenue). If small orders have high margins, the row-level average overweights them. The governed definition should be Sum(Profit)/Sum(Revenue) at the report level, with a documented filter for returns. Reason: a percentage is a ratio of aggregated amounts, not an average of per-row percentages.
Traps
- Calling a metric certified without a documented grain.
- Using a level metric to fix a filter problem.
- Letting report filters override a metric filter silently.
- Duplicating a metric under a new name, which breaks lineage.