The devil is in the detail. Aggregating data to a higher level can cause exceptions at a lower level to disappear.
For example, suppose a company would like to identify where it is giving discounts above a certain threshold. The following view shows that agriculture is the only industry exceeding the discount threshold. In absolute terms, this represents such a small number of units that it can be easily disregarded. However, it would be erroneous to conclude that there are no problems based on the summary view, as it is possible that individual customers within an industry are receiving a high discount rate.
The orange dots surface this detail, with the number of dots indicating the number of customers receiving discounts above the threshold. You can immediately see that there are three issues in the transportation industry, and clicking on a bar will filter to the customer detail view.
A Level of Detail Expression is required because the discount rate must be aggregated by customer, which is not in the visualization. More specifically, a fixed expression outputs a dimension that splits the orange dots into multiple points. If you are not convinced, change the expression to use include, and see what happens!