Bidirectional cross-filtering enables the filter context of a table on the many side of a relationship to flow to the one side of the relationship. Prior to support for bidirectional relationships, equivalent filter behavior commonly used in many-to-many relationships was only possible via DAX expressions. A common use case for bidirectional relationships is represented in the following data model:

In this model, a customer can have many accounts and an individual account can be associated with many customers. Given the many-to-many relationship between Customers and Accounts, a bridge table (CustomerAccount) is created that contains the combinations of customer and account key values. Due to the many-to one relationship between CustomerAccount and Accounts, a filter applied to the Customers table will only impact the Transactions fact table if bidirectional cross-filtering is enabled from CustomerAccount to Accounts. Without this bidirectional relationship, a filter applied to the Customers table would only impact the CustomerAccount table as single-direction relationships only flow from the one side of the relationship to the many. 

The only bidirectional relationships defined in the dataset for this project were described in the Bridge table section earlier in this chapter. Although powerful, and preferable for certain use cases, such as with bridge tables, bidirectional relationships can lead to unexpected or undesired query results. Additionally, the DAX CROSSFILTER() function makes it possible to selectively implement bidirectional relationship behavior for specific measures.