Answer

Additional Information

• Dimensions on the Rows or Columns shelf are treated hierarchically, so sorting on a dimension will not affect any dimensions left of it on the shelf. This is why [Sort By] must be in the first position on the Rows shelf.
• The RANK() function is a table calculation, which can be customized by changing how it is computed. For more information, please see Transform Values with Table Calculations
• If sort dimensions are not added to the view, then the sort may not work as expected. If the example view only had [Region] then sorting by "Month" would not appear to do everything because every region has the same months.
• If desired, you can create a calculated field that will change the granularity in the view. Something like:
  • CASE [Sort By]
    WHEN 'Region' THEN [Region]
    WHEN 'Month' THEN STR(DATETRUNC('month', [Order Date]))
    ELSE ' '
    END
• Dates need to be converted to an integer using INT() so that dates will retain chronological order
• "Month, Sales" is a nested sort. In general to create nested sorts, multiply the rank of the first sort by a large number and then add the rank of the second sort. The number should be large enough that the ranks 'stay separate' within the result. For example, adding together rank 1 and rank 3 should look like 103 rather than 4. This allows reach dimension/measure to be sorted separately in the final view.