Table filtered with data from another table who has a relationship is not working

[ClementDavid]

I have a visual showing the status of tickets (=issues). This comes from one table.

I have another table containing "labels". The tickets can have 0 to X labels.
The relationship between the two tables is therefore '1 to many' : 1 ticket can be linked to 0, 1 or many labels. 

I am now filtering my visual to get rid of any ticket who has the label "contractUpdate" in the Labels table. 
The filter shows the right amount of tickets in scope: 20 tickets. 

However, when I do apply the filter, more than 164 tickets are taken away from the visual, instead of the 20 expected:

If I change the cross filter direction from "both" to "Single" stricly nothing happens when I remove the label or add it. 
I am now really puzzled by this result to the point I think it's a bug.
Any thoughts? 
Your help will be very much welcomed. 
Thanks for reading! 

 

[ClementDavid]

How can we report this issue to Microsoft PBI team?
@Microsoft  @PowerBI 

[ClementDavid]

It seems to be a problem of blank/multiple possible values. 
The filter behaves correctly if and only if:

- all ticket KEYS are found in the table "Labels"

- there are only unique tickets KEYS in the table "Labels". 

 Which is an abnormal behaviour. Is Microsoft Power BI dev team working on this issue? 
Anyone knows?
I have seen similar questions on the net for this problem. 

[ClementDavid]

Thanks. I have used the formula above with COUNTDISTINCT but the formula you just provide gives the same result: 20 values are counted. Not 164+ ... 

[Bibiano_Geraldo]

Can you share the sample project without sensitive data?

[ClementDavid]

so I made a mock-up powerBI report file to show the problem, but it seems these message do not accept attached file and I dont have access to sharing websites through my work. 
But here it is: I have the exact same problem with this mock-up.
Filtering based on the contractUpdate label removes way too many tickets (way more than the 21 listed). 
It seems that it has to do with empty/blank values but I am not sure what/how. 

Still, that looks like a nasty bug from power BI. 

[ClementDavid]

I am afraid not 😞

each row contains something that could be considered as sensitive 

[Bibiano_Geraldo]

Hi,

Did you tried to use slicer to filter to see the result if keep the same?

[ClementDavid]

Hi, 
Yes I tried to use a slicer on the label. 
Gives exactly the same results. 

Now I have just noticed one thing: my version of Power BI is largely outdated. 
I will ask our support to update it and see if that solves the issue. I have good hope. 

[Bibiano_Geraldo]

Its always good to have Power Bi Updated, but its suppose to work.
 Now, what you can do to troubleshoot, its creating a measure to count the Label name where its equal to ContractUpdate, and then add that measure to a card to see what value appear.

the measure should look like this:

test = 
CALCULATE(COUNT([Label Name], FILTER([YOUR_TABLE, [LABEL NAME] = "contractUpdate"))

[ClementDavid]

Thanks again. 
I think I managed to do what you meant with the following formula: 

test = CALCULATE(DISTINCTCOUNT(Issues[ISSUE_KEY]), FILTER(Labels, [Label name] = "contractUpdate"))

This measure is placed in the "Issues" table. 

The result it gives is correct: 20. 
It does not explain the 164 of difference.

[ClementDavid]

Thank you for your feedback. 

I have tested with the new version of Power BI and same result/problem. 

However the formula you provide does not work. Count does not accept more than 1 argument. 
Can you please explicit what you attend to measure with it? 

[Bibiano_Geraldo]

You're right, i forgot to close parenteses in count function, try the bellow expression.

If you can attach the sample .pbix file its will be great, NOTE, with no sensitive data.

Thank you

test = 
CALCULATE(COUNT([Label Name]), FILTER([YOUR_TABLE, [LABEL NAME] = "contractUpdate"))