Relationships between fact tables and date dimension tables should always use single-direction cross filtering. If bidirectional cross filtering is used with date dimension tables, then filter selections of other dimension tables related to the given fact table, such as Promotion or Product, will reduce the date table rows available for date intelligence calculations. Similar to the example with shared dimensions, although this adjusted filter context is technically correct, it will often produce unexpected or undesired results, such as only the dates in which internet sales transactions were associated with a specific promotion type.
Note that the bidirectional relationship with the Date dimension table in this dataset is ...
