This other software applies a function to vary the value of the data, if the result set could allow someone to be identified.
What we need to do is to not allow a result of 1, when that actual result would be 1. In the above case, there would be only one person who matches the criteria born in China.
The FACT table is about 50 columns wide and 120M records.
The problem is that the total should be the total of the values displayed in the table, not the real total.
Also, we actually need to alter the total, so that someone cannot 'reverse engineer' a result, by working out the total in a set and then the total in a set with the exclusion of the one record which matches the criteria.
The capability of being able to assign custom formulas to various total cells on a pivot/straight tables is a very appreciated capability.
One side-effect I've observed from this technique is sometimes this will detach the dynamic re-sort capabilities of a straight table (on all columns, and not just the one with the "overridden totals"). So after I enter an expression that does this, then go to the straight table and click on the headers to apply/reverse sort, nothing happens.
The technique is really useful, but is there a way to sidestep impacting sort, and preserve both? ~E