Forum Discussion

stfox's avatar
stfox
Helper I
10 years ago
Solved

Cohort Analysis - DAX

Hi Folks

I want to create a Cohort Analysis. This involves calculating the the percentage of a cohort that is left in each of the subsequent months  following an initial signup. The screenshot below demonstrates the basic concept.  I want to avoild hard coding any date filters as a new cohort is added each month. 

 

I am thinking the solution involves a CALCULATE formula, but I am getting tripped up with the DAX to identify the denominator in the % calculation (the starting number of customers in each Cohort (40 in the first cohort, 35 in the second 30  in the 3rd cohort).

Can anyone help me out with the DAX?

 

 

Any help is much appreciated.  Excel/ PowerPivot File attached. 

 

Thanks Steve 

 

PowerPivot file

  • Hi Steve,

     

    Here are a couple of ideas I can think of depending on the logic you want to apply (they give the same answer using your sample data).

     

    These are additive over signup month (which you may want) and invoice month (which you may not want). You can modify that behaviour if needed.

     

    1. Denominator = For each Signup Month, Number of customers where Invoice Month = Signup Month

    =
    DIVIDE (
        [Count of Customer ID],
        CALCULATE (
            [Count of Customer ID],
            GENERATE (
                VALUES ( Table1[Signup Month] ),
                FILTER (
                    ALL ( Table1[Invoice_Month] ),
                    Table1[Invoice_Month] = Table1[Signup Month]
                )
            )
        )
    )

     

    2. Denominator = For each Signup Month, Number of customers in first Invoice Month with nonblank customers

    =
    DIVIDE (
        [Count of Customer ID],
        CALCULATE (
            [Count of Customer ID],
            GENERATE (
                VALUES ( Table1[Signup Month] ),
                FIRSTNONBLANK ( ALL ( Table1[Invoice_Month] ), [Count of Customer ID] )
            )
        )
    )

    Owen :)

5 Replies

  • Hi Steve,

     

    Here are a couple of ideas I can think of depending on the logic you want to apply (they give the same answer using your sample data).

     

    These are additive over signup month (which you may want) and invoice month (which you may not want). You can modify that behaviour if needed.

     

    1. Denominator = For each Signup Month, Number of customers where Invoice Month = Signup Month

    =
    DIVIDE (
        [Count of Customer ID],
        CALCULATE (
            [Count of Customer ID],
            GENERATE (
                VALUES ( Table1[Signup Month] ),
                FILTER (
                    ALL ( Table1[Invoice_Month] ),
                    Table1[Invoice_Month] = Table1[Signup Month]
                )
            )
        )
    )

     

    2. Denominator = For each Signup Month, Number of customers in first Invoice Month with nonblank customers

    =
    DIVIDE (
        [Count of Customer ID],
        CALCULATE (
            [Count of Customer ID],
            GENERATE (
                VALUES ( Table1[Signup Month] ),
                FIRSTNONBLANK ( ALL ( Table1[Invoice_Month] ), [Count of Customer ID] )
            )
        )
    )

    Owen :)

    • stfox's avatar
      stfox
      Helper I

      Excellent! Thanks Owen. Just what I was after. 

      • Anonymous's avatar
        Anonymous
        Not applicable

        Hi stfox OwenAuger 

        would you mind to share the solution in pbix format? Because these solutions doesn't work in my model. What data do you use for the matrix's column and row header?

    • GershwinMunich's avatar
      GershwinMunich
      Regular Visitor

      this formula does not work. it says too few arguments are trying to pass the filter function