Aggregate function for a measure - Count or Sum

  • I have a cube for counting # of members in different ways. The measure is "Member Count", which is the count of records. I have two choices for creating the measure.

    Choice 1: In the fact table, create an integer column MemberCounter with default value as 1, and create the measure as the Sum() for MemberCounter;

    Choice 2: just create the measure as the Count() for any non-null column (like the Date column).

    I wonder which one is better performance-wise.

  • seaport (6/12/2015)


    I have a cube for counting # of members in different ways. The measure is "Member Count", which is the count of records. I have two choices for creating the measure.

    Choice 1: In the fact table, create an integer column MemberCounter with default value as 1, and create the measure as the Sum() for MemberCounter;

    Choice 2: just create the measure as the Count() for any non-null column (like the Date column).

    I wonder which one is better performance-wise.

    I don't think you'll see much difference performance-wise between count() and sum(). For your purposes however, you may need to count unique members only (depending on your fact table's grain)...and Option 1 would not work in that case.

  • Martin,

    Thanks for the response.

    I do not think there is much performance difference between count() and sum() either.

    I do need an additional measure - Distinct Count of Member ID. If I understand correctly, with either a Sum() measure or a Count() measure in the cube, this measure can be created. Some people suggest putting Distinct Count in its own measure group for performance reason.

Viewing 3 posts - 1 through 2 (of 2 total)

You must be logged in to reply to this topic. Login to reply