Dimensions are categorizations of data, and the categorizations reflect how a business analyst wants to analyze data. When analysts say they want to see numbers "by" something or "over" something, they are identifying the dimensions of the data. Common dimensions are geography, product, customer, and time.
For KPIs, you can select any attribute column from the subject area as a dimension. When you design the KPI, you can either pin a specific data value to one or more dimensions (for example, Region = Central), or you can specify that you want all or some dimension values pinned when the KPI is added to a watchlist or scorecard and when a watchlist or scorecard is added to a dashboard. Pinning filters the data that the user sees, and after a value is pinned by the designer, the user cannot change the value.
When adding a KPI without a pre-pinned value to a watchlist, the designer can pin a specific value or specify a session or repository variable to set the dimension's value. If the designer does not pin a value or variable to a dimension, then the KPI user can select a value at runtime from the point of view area in either the KPI watchlist, from the scorecard that contains the KPI, from the variable prompt on a dashboard, or from the column prompt on the dashboard.
If a KPI with dimensions is output to an analysis, then the dimensions are displayed as columns and drills. If the dimension values are pinned, then the analysis is limited to the data that is determined by the pinned values.