Ranking with multiple rows and drilldown

[VBLOT]

Hello all,

I tried for many hours and differents ways but I can't manage to rank my matrix.

I have two lines in my Matrix :

As you can see in the screen bellow, I want to rank REF (higher level) separately from Line Designation (sublevel)

With ISINSCOPE function, Im able to see that to calculate REF ranking I need to have REFInScope = True && LineInScope = False

But with that in mind I somehow can't manage to do the right calculation (Ranking works when I have only 1 line)

See one of the DAX measure I've done :

Rank = 
VAR IsLineInScope = ISINSCOPE('Invoices List'[Line Designation])
VAR IsREFInScope = ISINSCOPE('Invoices List'[REF])

RETURN
    IF(
        IsLineInScope && IsREFInScope,
        RANKX(ALLSELECTED('Invoices List'[Line Designation]), [Total Samples collected]),
        IF(
            NOT IsLineInScope && IsREFInScope,
            RANKX(ALLSELECTED('Invoices List'[REF]), [Total Samples collected]),
            BLANK()
        )
    )

Can you please help me? I've tried like 50 differents measures, watch videos and nothing works. I also tried with RANK and not RANKX....

[OwenAuger]

Hi again @VBLOT 

Could you share a sanitised PBIX file that exhibits the problem? Even filtered down to a couple of REF values.

(share a link to Google Drive / OneDrive etc).

From what you've posted, it should be working, but there must be some particular feature of your data model that is causing this.

[VBLOT]

Hi again @OwenAuger,

Please find a sanitised file here :
https://norishare.com/P9TDkeS9yJF

Thanks for your help.

[OwenAuger · Accepted solution]

Hi again @VBLOT 

Thanks so much for the file! That made it a lot easier to debug.

It seems that the problem actually relates to how the visual level filter "Sum of Nb samples collected > 0" interacts with the ALLSELECTED () modifier used to produce the table for ranking.

A safer way to write the measure is below (see attached PBIX). Actually in their Whitepaper, SQLBI recommend this general approach of adding column(s) to the "relation" argument of window functions.

I might have to get back to you with an explanation of why it works when I've analyzed further.

Rank FIX = 
VAR IsLineInScope = ISINSCOPE ( 'Invoices List'[Line Designation] )
VAR IsREFInScope = ISINSCOPE ( 'Invoices List'[REF] )
VAR Result =
    SWITCH (
        TRUE ( ),
        IsLineInScope && IsREFInScope,
            RANK (
                DENSE,
                CALCULATETABLE (
                    ADDCOLUMNS (
                        SUMMARIZE (
                            'Invoices List',
                            'Invoices List'[Line Designation],
                            'Invoices List'[REF]
                        ),
                        "@TotalSamples", [Total Samples collected]
                    ),
                    ALLSELECTED ()
                ),
                ORDERBY ( [@TotalSamples], DESC ), ,
                PARTITIONBY('Invoices List'[REF])
               
            ),
        NOT IsLineInScope && IsREFInScope,
            RANK (
                DENSE,
                CALCULATETABLE (
                    ADDCOLUMNS (
                        SUMMARIZE ( 'Invoices List', 'Invoices List'[REF] ),
                        "@TotalSamples", [Total Samples collected]
                    ),
                    ALLSELECTED ( )
                ),
                ORDERBY ( [@TotalSamples], DESC )
            )
    )
RETURN
    Result

Regards

[OwenAuger]

Hi @VBLOT 

Your logic is essentially correct!

In fact your measure works fine in a test PBIX I created, as long as [Total Samples Collected] always has positive values (example in attached PBIX).

However, since it wasn't working correctly for you, it's possible that there is a complication in your model. Are there any sort-by columns defined by any chance?

A more robust way I would suggest writing this measure is using the RANK function, since we can specify partitioning.

Rank = 
VAR IsLineInScope =
    ISINSCOPE ( 'Invoices List'[Line Designation] )
VAR IsREFInScope = ISINSCOPE ( 'Invoices List'[REF] )
VAR Result =
    SWITCH (
        TRUE ( ),
        IsLineInScope && IsREFInScope,
            RANK (
                DENSE,
                CALCULATETABLE (
                    SUMMARIZE (
                        'Invoices List',
                        'Invoices List'[Line Designation],
                        'Invoices List'[REF]
                    ),
                    ALLSELECTED ( )
                ),
                ORDERBY ( [Total Samples Collected], DESC ),
                PARTITIONBY ( 'Invoices List'[REF] )
            ),
        NOT IsLineInScope && IsREFInScope,
            RANK (
                DENSE,
                CALCULATETABLE (
                    SUMMARIZE ( 'Invoices List', 'Invoices List'[REF] ),
                    ALLSELECTED ( )
                ),
                ORDERBY ( [Total Samples Collected], DESC )
            )
    )
RETURN
    Result

PBIX attached for reference

• Rank Original is your posted measure.
• Rank is updated measures.

Regards

[VBLOT]

Hello @OwenAuger !

Thanks for your answer ! It is indeed working in the PBI you linked but sadly it doesn't work on mine.

I've checked and I don't have any sort-by columns in Invoices List, the code in entering conditions as expected but somehow, the rank part in the REF scope isn't working and I still have 1 everywhere...

I also know that I have blanks in my Nb samples collected. But it shouldn't have any impacts as I filtered the result to have only those greater than 0 and in my Total Samples collected measure, I make sure that I don't consider BLANK().

Total Samples collected = 
    CALCULATE(
        SUM('Invoices List'[Nb samples collected]),
        FILTER('Invoices List','Invoices List'[Nb samples collected] <> BLANK())
    )

Do you have another idea ?