Back to Basics: what’s the difference among dimensions, measures and attributes?

 

Originally Posted 10/19/2010 

I keep going back to the basics because I was asked (thanks Mary!) to explain in the same way the difference among dimensions, measures an attributes. Actually this was not from a BI professional but from the comptroller of a medium sized company which is setting up its first organized BI system. She's been told about but she's not sure of having a clear picture. So I'm trying to help.

To understand the difference among dimensions and measures I usually tell a very short story. 
Let's suppose that Mary's boss drops by her office and says "Mary, I need to know the sales amount". Mary is a bit surprised and replies "The sales amount of what? And in which time span?".

"The sales amount" is a measure.

"what" and "time span" are the dimensions

.The sales amount alone is a lousy concept, it does not mean anything, if you do not better qualify it with a period, a group of customers, a group of products etc. The measure is a number which tells us something about the company, the dimension qualifies it and "slices" it according to relevant business entities. Measures can be summarized or used for math calculations, dimension cannot.

I can hear the objections "What? My area sales are the sum of all the salesmen working in the area! I sum the salesmen!". No, you do not sum salesmen, you sum the sales amount of each salesman. You still sum a measure.

The attribute

is a tag associated to the dimension. It is something you would never use to slice data. For example, the customer's receptionist name. Would you ever slice your sales to know how much you sold to companies whose receptionist is named Charlene? Receptionist's name is a useful piece of information to keep at hand, but it's not relevant for analysis.

From a technical perspective, it is more convenient labeling it as an attribute rather than adding a useless dimension.

I hope this simple explanation is clear, but, as usual, I'm open to even clearer explanations.

A vous!

 

  • Mar  3 2011,  5:01 PM

    Jaime Fitzgerald responded:

    You did it.  A great, simple, clear explanation of the differences!  For anyone learning to think about data, I will be referring them to this link! Thanks!