Distribution of counts

[pieterhkruger]

Hi,

I would like to calculate the distriution of the number of occurences of a field.  More specifically, I want to see the distribution of the number of applications customers have in a particular time period.

For example, I can easily for the first column (customer number) determine how many applications they made:

However, I want to be able to create a graph / summary table that would tell me:

Number of applications	Number of customers (having this number of applications)
1	11
2	1

How would I go about doing so with DAX (and without creating a table that would not filter according to filters on a page)?

[CheenuSing]

Hi @pieterhkruger 

1. Create a measure Distribution = COUNT(yourtablename[Customer])

2. Now create a table visual with

   YourTable[Count] & Distribution as values.

   For YourTable[Count] value set to do not summarize.

CheenuSing

[pieterhkruger]

Hi,

Thanks for the response.

I don't exactly understand how to do the second step.  Do I have to use the SUMMARIZE statement?  But in that case I cannot use a calculated field as the second parameter.  Could you perhaps provide an example?

Thanks

[CheenuSing]

All you need is create the measure as at step 1.

For step 2 it is the table visual from visualisation panel.

CheenuSing

[pieterhkruger]

Sorry, maybe I understand your solution incorrectly, but let me explain with an example:

If I have the following data:

ACCOUNT_NUMBER	CUSTOMER_NUMBER
1	A
2	B
3	C
4	D
5	E
6	E
7	F

The creating a measure - COUNT([CUSTOMER_NUMBER]) and creating a visual table based on that, would give me:

CUSTOMER_NUMBER	COUNT(Customer_number)
A	1
B	1
C	1
D	1
E	2
F	1

However, what I want, requires another step, because what I now want is this:

COUNT(Customer_number)	COUNT(COUNT(CUSTOMER_NUMBER))
1	5
2	1

[OwenAuger · Accepted solution]

@pieterhkruger 

To make this work, I believe you'll need to create a disconnected dimension table containing all possible "count" values.

Then create a Frequency measure to count the number of CUSTOMER_NUMBERs with a given count.

See for example this pbix using your sample data

I created a table called 'Count' and a measure Frequency

Count = 
VAR MaxCount = 
    MAXX ( 
        VALUES ( YourTable[CUSTOMER_NUMBER] ),
        CALCULATE ( COUNTROWS ( YourTable ) )
    )
RETURN
    SELECTCOLUMNS (
        GENERATESERIES ( 0, MaxCount ),
        "Count", [Value]
    )

Frequency = 
SUMX ( 
    'Count',
    COUNTROWS ( 
        FILTER ( 
            VALUES ( YourTable[CUSTOMER_NUMBER] ),
             CALCULATE ( COUNTROWS ( YourTable ) ) = 'Count'[Count]
        )
    )
)

The Frequency measure uses SUMX to iterate over the 'Count' table so it can aggregate over multiple Count values.

You can then create visuals showing Frequency filtered by 'Count'[Count]

Regards,

Owen

[pieterhkruger]

Thank you very much!

This works well, especially for showing the distribution over the whole range of possible counts - which can be beneficial.

(I also got it to work with a COUNT and SUMMARIZE statement, together with a bi-directional join between the 2 tables, in which case it only keeps the range of counts relevant for the selection).