Inconsistent filter context behavior in matrix visualization

[doug_morris]

In a matrix visualization, a cell which is the intersection of table columns A and B does not manifest filter context consistently as you change how A and B are configured in the matrix on Rows and Columns.

This appears to be a defect or at best a very confusing feature. It makes the matrix visualization considerably less functional and more likely to show erroneous/unexpected results.

Here is a very simple example. The inconsistent behavior is not specific to the example shown here. It can be reproduced with any comparable dataset.

1. Create a table.

Table = DATATABLE(

    "Size", STRING, "Color", STRING, "Quantity", INTEGER,

    {

        {"Large", "Green", 1},

        {"Large", "Yellow", 2},

        {"Small", "Green", 3},

        {"Small", "Yellow", 4}

    })

2. Create a matrix visualization. 

• Size and Color on Rows.
• Sum of Quantity on Values.
• Expand all down one level in the hierarchy

3. Filter on Sum of Quantity is less than 4.

4. Move Color to Columns. Notice that we do not have the same Values as in step #3.

[parry2k]

@doug_morris it's ok. Sometime we miss things. And I have seen similar question many times and that's why I was quick to share the posts. Anyhow, glad you know what is going on. We all need to be slow down and read things bit  more carefully. I also miss things, and many times.

Best!

[parry2k]

@doug_morris appreciate the feedback. If you don't want to read the article and understand why it is happening, not sure what else can be done. I don't want to explain something which is already well explained.

Good luck!

[doug_morris]

My comment about it being spam was regarding your other post which gave no solution but only an advertisement for your YT channel. Maybe that post was an mistake. I didn't see your second response or overlooked it because it appeared so much like the other post. I see the linked article in your second response and thank you for it. That addresses my question directly. I will add a reply to that post, give it a thumbs up, and accept it as a solution.

[parry2k]

@doug_morris and here 

[doug_morris]

To be frank, your response is spam. If I could give a thumbs down, I would.

[parry2k · Accepted solution]

@doug_morris I think you are using a visual level filter and it is very well explained here. It is all about understanding how DAX works. 

[doug_morris]

Thank you, this article addresses the issue. As it says in the article, "For a matrix visual, the filter granularity is defined by attributes applied to the Rows section of the matrix, ignoring the attributes applied to the Columns section." I think it's extremely non-intuitive for the matrix visualization to ignore Columns from the filter granularity. Simply changing row/column format without any change to the aggregation in a viz, I would not expect the effect of a filter to change. I guess the workarounds are to move the filter to a page-level filter (which won't work for measures) or to move the filtering into a DAX expression. Both seem to be muddled workarounds, but I guess that's what we have to live with given the design. Maybe there are good reasons the matrix viz was designed that way, but it's not clear to me.