Forum Discussion

yamacha's avatar
yamacha
Icon for Helper I rankHelper I
2 years ago

Matrix calculation with multi-layered SUMX

Hi there

I have been struggling with the  Matrix Calculation for years, and SUMX + VALUES seems to be most handy way that can cover most situations. However as you can  below it won't help when there are more than 2 layers of context (i.e. LP_ABBR (customer name), MonthDate and ProductLine5 here)... The displayed outcome are obviously wrong... Need your help on it. Thanks!

 

 

 

 

Need Help

9 Replies

  • ahadkarimi's avatar
    ahadkarimi
    Icon for Solution Specialist rankSolution Specialist
    2 years ago

    Hi yamacha, try this SUMMARIZE, if you encounter any issues, let me know.

     

    ERROR_LP_PLS = 
SUMX(
    SUMMARIZE(
        ACTUAL_LP,
        LP_ABBR[LP_ABBR],
        TIME[MonthDate],
        MASTER_PL[ProductLine5],
        "ActualQty", SUM(ACTUAL_LP[Actual Qty]),
        "ForecastQty", SUM(FCST_LP[FCST Qty])
    ),
    ABS([ActualQty] - [ForecastQty])
)

     

     

    Did I answer your question?  If so, please mark my post as the solution!✔️
    Your Kudos are much appreciated!  Proud to be a Responsive Resident!

    • yamacha's avatar
      yamacha
      Icon for Helper I rankHelper I
      2 years ago

      Thanks ahadkarimi

      I created two columns (end with Neo) with your codes as below. As only 'ACTUAL_LP' is summarized in your code so it will miss the lines w/o actual demand but still with forecast, and ultimately underestimate the error.

       

  • Anonymous's avatar
    Anonymous
    Not applicable
    2 years ago

    Hi All,

    Firstly  ahadkarimi thank you for your solution!
    And yamacha to my understanding,you want to get the error for each product line for that month and then add that up, right? 
    The matrix can be aggregated for you, you don't need to write your own aggregation function, maybe my data is too simple to reproduce your problem, if you can can you upload some sample data, maybe it will help to solve your problem.

     

    ERROR_LP = ABS(SUM('basic data'[Actual Qty])-SUM('basic data'[FCST Qty]))

     

    I would be extremely proud if I could solve your problem and look forward to hearing from you!

    Hope it helps!

     

    Best regards,
    Community Support Team_ Tom Shen

     

    If this post helps then please consider Accept it as the solution to help the other members find it more quickly.

     

    • yamacha's avatar
      yamacha
      Icon for Helper I rankHelper I
      2 years ago

      Hi Xingshen

      Thanks for the comments, but unfortunately it's not that simple. There is a further layer below the product line that is the UPN (SKU), and I need to calculate the error on these 2 different levels (that's why you see I have to columns for Error). 

      • Anonymous's avatar
        Anonymous
        Not applicable
        2 years ago

        Hi yamacha ,
        I apologise for not providing you with a suitable solution as I didn't understand you correctly.
        If sumx and value may not be able to calculate the value of the difference between the strata correctly, you can use the alexecpt function to add a qualification to determine the stratum you need to calculate

        -- ERROR_LP_PL5 = 
CALCULATE(
    MAXX(
        VALUES('Table'[ProductLine5]),
        ABS(SUM('Table'[Actual Qty]) - SUM('Table'[FCST Qty]))
    ),
    ALLEXCEPT('Table', 'Table'[LP_ABBR], 'Table'[MonthDate], 'Table'[ProductLine5])
)

        -- ERROR_LP_UPN = 
CALCULATE(
    MAXX(
        VALUES('Table'[UPN]),
        ABS(SUM('Table'[Actual Qty]) - SUM('Table'[FCST Qty]))
    ),
    ALLEXCEPT('Table', 'Table'[LP_ABBR], 'Table'[MonthDate], 'Table'[ProductLine5],'Table'[UPN])
)

        If in doubt, check out the pbix file I uploaded and I'd be honoured if I could help you out!

        Best regards,
        Community Support Team_ Tom Shen

         

        If this post helps then please consider Accept it as the solution to help the other members find it more quickly.