Aggregation determines how a source field is summarized for a query. Set the reusable default in the Analysis Model; use a report override only when one component needs a different valid summary.

Set the model default

Select a regular Measure in the model tree, then choose its Aggregation type in the Attributes panel and save the model.

The current model designer provides:

• Sum, Average, Min, and Max
• Count and Distinct count
• Std dev (population) and Std dev (sample)
• Variance (population) and Variance (sample)

When a source field is first set as a Measure, a numeric field starts with Sum and a nonnumeric field starts with Count. Fact Count is fixed to Count. Calculated measures use a formula and do not have an aggregation selector.

Choose the method from the field's business meaning. In particular, do not Sum unit prices, rates, balances, or other non-additive values merely because they are numeric. Database support for statistical aggregations can vary, so test the result against the source system.

For the complete model workflow, see Measures and Calculated Measures.

Override aggregation in a report

For a Measure field already assigned to a component:

1. Open the field's menu.
2. Select Aggregation type setting.
3. Choose an enabled method, or choose the default option to remove the override.
4. Save and validate the report.

The report menu can list Avg, Sum, Max, Min, Count, Distinct count, Median, population or sample standard deviation, and population or sample variance. Only methods allowed by that field's model metadata are enabled.

A report override affects that field instance in the component query. It does not change the model default or other reports.

Validate the result

• Confirm the aggregation matches the table grain and the Measure definition.
• Check relationship cardinality before investigating an unexpected total; duplicated join rows can make a correct Sum look wrong.
• Compare a representative result with a trusted source query.
• Use a calculated measure for ratios or weighted calculations instead of averaging already aggregated percentages.
• Prefer the model default when the same rule should apply everywhere.