Appearance
Creating an Analysis Model
Before you start, make sure you have permission to create Analysis Models and can read the required data source schema and tables.
1. Start a model
Open Data > Models, then click New.
The new model opens without tables. Select a Datasource and Schema in the Data panel, or click Choose datasource on the canvas.
2. Add tables
Find a table in the Data panel, then use one of these methods:
- Click the add control beside the table.
- Double-click the table.
- Drag the table onto the canvas.
Adding a table initially creates:
- A Dimension containing its fields as Attributes.
- A Measure Group containing eligible numeric fields as Sum measures and a Fact Count measure.
Primary keys, foreign keys, and identifier-like numeric fields are excluded from automatic Sum measures. These generated objects are a starting point; remove objects that do not match the table's real business role.
3. Create relationships
Connect tables using their actual business keys. Set Key values are unique only when the selected field combination is unique on that side of the relationship.
For details, see Establishing Table Relationships.
4. Define the semantic model
Use the Analysis model tree and Attributes panel to:
- Keep descriptive fields in Dimensions.
- Set the correct aggregation for each Measure.
- Add Hierarchies only where their level order has a clear analytical meaning.
- Add descriptions, aliases, semantic roles, units, and time settings.
- Create calculated measures when the calculation must be reusable across reports.
5. Review and save
If the Insights panel is displayed, expand Model diagnostics and review its findings before saving. Insights is hidden when there are no diagnostics or metric bindings.
- Warnings and hints do not stop the save operation.
- Errors require an explicit decision to return to the model or continue saving.
- Server-side model verification runs after the client diagnostics.
- A model with no tables cannot be saved.
Click Save. On the first save, provide the analysis model name.
Use Save as when you need a separate model rather than changing the current one.
Continuing after a diagnostic error can save an incomplete or inconsistent model. Resolve structural and reference errors whenever possible.