Settings: Modeling
The Modeling tab controls defaults applied when creating models, columns, and date/time fields, as well as custom format definitions and Model Health scoring configuration.
Allow Publish with Validation Errors
Pre-publish schema validation blocks publishing when the data source is missing tables or columns, or when types don't match (processing would fail on the server). Turn this on to bypass validation and publish anyway.
Off by default.
Default Model Culture
The locale applied to new models when they are created (e.g., en-US - English (US)). This affects how the model interprets number and date formatting by default.
Default Column Summarize By
The default summarization behavior applied when creating new regular columns. Set to Do not summarize to prevent columns from auto-aggregating in client tools like Power BI.
Default Calculated Column Summarize By
Same as above, but applied when creating new calculated columns. Independent from the regular column default so you can set different behaviors for each.
Default Date Format
The format string applied to new date/time columns on import and Refresh Schema. Pick None to leave new date/time columns as General format.
Custom Number Formats
Named number format entries that appear in format pickers throughout the app. Each entry has a name and a format string.
Click + Add Number Format to create one. For example:
| Name | Format String |
|---|---|
| US Whole Number | #,##0 |
| Percentage | 0.00% |
| Currency (USD) | $#,##0.00 |
Custom Date Formats
Named date format entries that appear in format pickers throughout the app.
Click + Add Date Format to create one. For example:
| Name | Format String |
|---|---|
| US Date Format | MM-dd-yyyy |
| ISO Date | yyyy-MM-dd |
Model Health
These settings control how Model Health behaves and how scores are calculated.
Auto-compute on Model Open (Working Model)
When enabled, Model Health automatically computes scores whenever you open a working model. Turn this off if you prefer to run health checks manually.
Show Readiness Dots in the Explorer
When enabled, colored dots appear next to objects in the Explorer panel indicating their readiness status at a glance.
AI Readiness Score Weights
The AI Readiness score measures how well-prepared a model is for AI consumption (MCP servers, Copilot, and other AI tools). Each weight controls what percentage of the total score that category represents. Weights must sum to 100%.
| Category | What It Measures | Default |
|---|---|---|
| Descriptions | Tables, columns, and measures have descriptions | 40 |
| No M:M | Absence of many-to-many relationships | 15 |
| No bidirectional | Absence of bidirectional cross-filtering | 15 |
| No bridge tables | Absence of bridge tables | 15 |
| Synonyms | Entities have synonyms defined in an ontology | 10 |
| Date table | A proper date table is present | 5 |
Increase a weight to make that category matter more to the score. For example, if descriptions are critical for your AI workflows, keep the Descriptions weight high.
Model Quality Score Weights
The Model Quality score measures how well a model follows modeling best practices. Weights must sum to 100%.
| Category | What It Measures | Default |
|---|---|---|
| Measure descriptions | Measures have descriptions | 25 |
| Naming clarity | Tables, columns, and measures follow clear naming conventions | 20 |
| Structure | Model structure follows star schema best practices | 20 |
| Relationships | Relationships are well-defined and follow best practices | 15 |
| Date table | A proper date table is present | 15 |
| Format strings | Measures have format strings applied | 3 |
| Display folders | Measures are organized into display folders | 2 |
Grain Mismatch Threshold
The minimum percentage of fact tables a dimension must relate to before Model Health flags it. Model Health looks at dimensions that connect to some but not all of your fact tables and calculates what percentage they cover. If that coverage falls below this threshold, it raises a "Fact table grain mismatch" warning.
At the default of 100, any dimension that doesn't relate to every fact table is flagged. Lower the threshold if your model intentionally has dimensions that only apply to a subset of fact tables - for example, a "Promotion" dimension that only relates to Sales but not Inventory.