You use both filters and selection steps to limit the results that are displayed when an analysis is run. This means that the results answer a particular question. Together with the columns that you select for an analysis, filters and selection steps determine what the results contain. Based on the filters and selection steps, only those results that match the criteria are shown. For example, depending on the industry in which you work, you can use filters and selection steps to learn who are the top ten performers You can also learn the dollar sales for a particular brand, and which are the most profitable customers.
Filters and selection steps are applied on a column-level basis and provide two methods for limiting the data in an analysis. A filter is always applied to a column before any selection steps are applied. Steps are applied in their specified order. Filters and selection steps differ in various ways:
Filters are applied directly to columns before the query is aggregated. Filters affect the query and thus the resulting values for measures. For example, suppose that you have a list of members in which the aggregate sums to 100. Over time, more members meet the filter criteria and are filtered in, which increases the aggregate sum to 200.
Selection steps are applied after the query is aggregated and affect only the members displayed, not the resulting aggregate values. For example, suppose that you have a list of members in which the aggregate sums to 100. If you remove one of the members using a selection step, then the aggregate sum remains at 100.
Selection steps are per column and cannot cross columns.
Another kind of filter, called a prompt, can apply to all items in a dashboard. Prompts can be used to complete selection steps and filters at runtime. For information, see Prompting in Analyses and Dashboards.
A prompted filter is a filter whose operator is set to is prompted. This operator is valid for a column that contains text, numbers, or dates. When you select the is prompted operator for a filter's column, you flag the column as ready to be filtered by a prompt. When a prompt is used, results include only records where the data in the column that is prompted matches the user's choices.
The is prompted operator is required for columns that are included in prompts where no prefiltered values are wanted.
For more information, see Specifying Values for Filters and Advanced Techniques: How Dashboard Prompts and Analysis Prompts Interact.