Custom Subtotals in Matrix for Different Categories

[Jansco]

I’m looking for a way to calculate custom totals based on a filtered Matix.

Let’s say I have 15 variables, including the following:

A

B

B/A

C

D

C/D

For single variables a normal total is fine. For B/A the total should actually be the total of B/the total of A.

I know I can create measures for each of these, summing the values or dividing, but is there any way to put all of the totals neatly in one column?

[d_gosbell]

If you have a table with the list of variables (which is not connected to any of your other tables) you could use this table and a measure with a SWITCH statement to bring all your values into the one column. The switch can either reference other measures or do the calculations inline.

eg

= SWITCH( SELECTEDVALUE( Variables[Variable]) 
 , "A",  <calc for A>

, "B",  <calc for B>

, "B/A",  <calc for B/A>

, <default calc>

)

[Jansco]

Okay.  So, I assume there is no way to do this if my list of variables IS connected to my other tables?

My table looks something like this

	January	March	April	May
Customer 1
A	10	20	30	40
B	20	30	40	50
B/A	2	1.5	1.3	1.25
Customer 2...

[d_gosbell]

So If you've already got the amounts calculated can't you just unpivot the data so that you have it in the following format

Customer	Variable	Month	Value
Customer 1	A	Jan	10
Customer 1	A	March	20
Customer 1	A	April	30
Customer 1	A	May	40
Customer 1	B	Jan	20
Customer 1	B	March	30
Customer 1	B	April	40
Customer 1	B	May	50
Customer 1	B/A	Jan	2
Customer 1	B/A	March	1.5
Customer 1	B/A	April	1.3
Customer 1	B/A	May	1.25

[Jansco]

Here is what I see as generated by Power BI subtotals:

	January	March	April	May	Totals
Customer 1
A	10	20	30	40	100
B	20	30	40	50	140
B/A	2	1.5	1.3	1.25	6.05 (wrong!)

But, this is what I want:

	January	March	April	May	Totals
Customer 1
A	10	20	30	40	100
B	20	30	40	50	140
B/A	2	1.5	1.3	1.25	1.4 (correct!)

Unpivoting the data would allow me to create these totals?

[d_gosbell · Accepted solution]

So if you unpivot and use the disconnected table it should work fine.

Note: I would hide the variable column in Table1 as the switch measure is based off the 'Variable' table. If you use the column from Table1 the measure will not work.

[atpbi10]

Hey thanks for this hint. It works nicely for one level matrix. How about if we have two level nested matrix? Like there are multiple variables for each of top level and I want to have custom subtotals for the top level variables but still be able to drill down and see the correspodning subvariables one level below.

I tried this solution but there are some weird things happening with the matrix table.

[d_gosbell]

---

@atpbi10 wrote:

Hey thanks for this hint. It works nicely for one level matrix. How about if we have two level nested matrix? Like there are multiple variables for each of top level and I want to have custom subtotals for the top level variables but still be able to drill down and see the correspodning subvariables one level below.

I tried this solution but there are some weird things happening with the matrix table.

---

It's hard to say as you have not provided any detail and I can't help fix "weird things" without details of what is going on. In theory you should be able to make this technique work across multiple levels, but it may get complicated an you may need to have multiple levels of nested switch statements. 

Can you provide a link to an example pbix file with some data the shows what you are trying to do? If you can also provide the expected out comes given the example data and how you want the calculations to work we can probably help you out. (see https://community.powerbi.com/t5/Community-Blog/How-to-Get-Your-Question-Answered-Quickly/ba-p/38490 )

[atpbi10]

So here is my example file. I have three levels in the matrix table. As you can see in my attempt using your solution it changes all the values for the given variable that I want to have custom total. Instead of whats happening my desired result would be to only apply the custom total on the highest level (level 1-which is "Variable" column). And if I drill down then the values would behave normally, so they would sum for the corresponding variable in the input file.

https://drive.google.com/file/d/1wC78oj6i9srmW7B0U2bUEIR7KqlNSV2N/view?usp=sharing

[d_gosbell]

As I suggested earlier you can do this by adding some nested logic. So inside the "Gross Profit" condition we can then check if there are any filters on the L2 or L3 tables (if they are not filtered then we are looking at a total value).

eg.

Measure =
SWITCH (
    SELECTEDVALUE ( 'Sheet1'[Variable] ),
    "Gross profit",
        SWITCH (
            TRUE (),
            NOT ( ISFILTERED ( VariableL2 ) || ISFILTERED ( VariableL3 ) ), 
            12,
            SUM ( 'Sheet1'[Value] )
        ),
    SUM ( Sheet1[Value] )
)

This produces the following output

[Sagejah9]

I tried this but it returns blank. 

[Anonymous]

Hi @d_gosbell ,
I tried this. Basically, I created a new table with the categories in one column ad used that in switch case. 
Thanks a ton for this solution.
Although, I got confused when you mentioned "unpivot" . 
I am still not sure why unpivot?

Regards,
Aishwarya

[d_gosbell]

So "Pivoting" is the act of taking an attribute (like the month) and creating separate columns by grouping on the values of that attribute (as in the original dataset you posted).

"Unpivotting" is the reverse of this operation where you take the grouped columns and transform them back into rows. This is what I did in my previous response to take your dataset where I transformed your dataset so that it had a single "month" column instead of a column for each month.