The Default Summarization property should be revised from Power BI's default setting to the Do not summarize value for all columns. Power BI will apply a Default Summarization setting of Sum for all columns with a numeric data type (whole number, fixed decimal number, decimal number) when a table is first loaded to the data model. As shown in the following screenshot, a summation symbol (∑) will appear next to the field name in the fields list if a Default Summarization other than Do not Summarize is enabled:

As illustrated in the previous image, the Default Summarization property for a column can be accessed via the Modeling tab of the Data View. Additionally, as with other metadata properties, Default Summarizaton can also be accessed from the Report View. As mentioned in the Data View section earlier, implementing metadata changes, such as Default Summarization and Data Category, via the Modeling tab from the Report View is the only option for DirectQuery models. 

If a user selects a column with Default Summarization enabled, the specific aggregation specified by the property (for example, Sum, Average) will be returned rather than the grouping behavior of Do not summarize. In many cases, the numeric column is only used to group measures, such as Internet Net Sales by Product Dealer Price, and DAX Measures can be written for any needed calculation logic. Additionally, Default Summarization can create confusion, such as when a user expects a sum aggregation based on the summation symbol but the model author has applied an alternative default summarization (for example, Minimum, Average). Alternatively, the names assigned to DAX measures, such as Average Product Dealer Price, make it clear which aggregation is being applied.

For these reasons, it's recommended to convert the default summarization setting to Do not Summarize. A broader concept of this recommendation is to build essential DAX Measure expressions into the dataset, as described in Chapter 4, Developing DAX Measures and Security Roles, to make Power BI datasets more flexible and powerful for users and report developers.