Subtract two table columns in Excel

Use a calculated field

The best way I've found to achieve this is to add a calculated field to the table that produces the correct sort order, and then to hide that column.

The sort order field needs to work by scaling each sort column up or down so that no column's bit of the output sort value overlaps with any other.

So I scale OrderCount up by a big number, and total Value down by a big number:

= (OrderCount*1000) + (Value/1000) 

For Jemaine's figures above, this will produce a value of 3000.18. Bret would be 3000.07, Murray/Brian 1000.09 etc - the correct order.

If I wanted to add a third sorting condition I'd have to scale that down by an even bigger number:

= (OrderCount*1000) + (Value/1000) + (3rdCondition/1000000)

Adding the Calculated field

To actually implement this in the pivot table, I need to add a column to my data table so that the calculated field can count orders. I've simply added an OrderCount column and set the value to 1 for every row:

Now we can add the calculated field. Select this option from the ribbon and enter the formula:

My pivot table now looks like this, sorted correctly.

Tidying Up

This SortOrder column might confuse other users, so I can just hide the whole column.

However, I often prefer not to hide columns (personal preference) so to make this column unobtrusive I can rename the column to (ie a blank space), and format the number column for the SortOrder values to a custom format of "";"";"". If Excel autoadjusts column widths this makes the column barely noticeable.

You can use INDEX to return the full row:

=SUM(INDEX(Table1,ROW($ZZ1),0))

Now as it fills down it will change which row it is returning.

We can change out ROW($ZZ1) for MATCH and not worry about order:

=SUM(INDEX(Table1,MATCH([@Item],Table1[Item],0),0))

As you can see neither refers to a column range (except the MATCH) allowing the addition of columns in Table1 and it will adjust automatically.