Data grain
State what one row represents.
Examples include one expense, one approval cycle, or one organisation-month.
Data modelling
Define the grain before the metric.
A reporting model controls what one row means and where business logic lives.
Visual model
From records to decision-ready metrics.
The grain fixes the meaning of each row. Facts and dimensions organise analysis. Marts and the semantic layer expose governed measures.
Data grainStar schemaData martsSemantic layer
InputSource records
→
MeaningDefine one row
→
ModelFacts + dimensions
→
ServeData marts
→
UseSemantic layer
Star schema
Separate events from context.
Facts hold measures. Dimensions provide time, organisation, vendor, and category context.
Data marts
Serve one decision area.
A mart gives users a stable view for spend, contracts, approvals, or service performance.
Semantic layer
Define the metric once.
The dashboard reads governed measures instead of rebuilding business logic.
Where I used it
Five reporting marts
dbt models serve spend, contract, vendor, approval, and currency analysis.
View project →Organisation-month reporting
A controlled model combines workforce, absence, demand, and delivery measures.
View project →