Counting subsets of rows

[Believer]

Considering the following sample data, where the first 4 columns are in a table and the last column (Cust_RecID_Count) is calculated:

Customer	RecID	SecID	PosID	Cust_RecID_Count
A	1	12	1	2
A	1	15	1	2
A	2	12	1	1
B	2	17	1	3
B	2	19	2	3
B	2	55	1	3
B	3	19	1	1
C	1	52	1	1

How would I calculate Cust_RecID_Count, the row count for each Customer/RecID pair?  Any filters which may be applied in the report page should not affect the value.  Every time Customer/RecID A/1 appears, the result should always be 2.  For B/2 it would be 3, and all others are 1.

I’ve been using Power BI for a while but have not done much with CALCULATE, been tearing my hair out trying to get it right.  Beginning to wonder whether a CALCULATE expression is even the way to go vs a completely different method, such as dynamically building a separate table of Customer/RecID/Count?  Performance Is fairly important since my source table has over 12 million records.

[AllisonKennedy]

@Believer 

If performance is important you should do this as a Measure, not a column.

Try this: 

Cust_RecID_Count = CALCULATE([Customer Count], ALLEXCEPT('Table', 'Table'[Customer  ], 'Table'[RecID  ]))

[Believer]

@AllisonKennedy ,

Thanks for the solution. The good news is it works perfectly based on what I told you, and I was able to get it working on my production data. The bad news is I discovered that there’s another table and relationship involved, and when I add that in it makes a mess of the results, adding in a lot of rows which shouldn’t be there. Although I can filter the visual to remove the blank “Customer Count” that strikes me as a hack.
I added the Customers table to the model, relationship, etc. Here are the screen shots, not sure why I cannot upload the updated PBIX file here... I get an error "The file type (.pbix) is not supported."

[AllisonKennedy]

@Believer  Re uploading .pbix files - you need to be a superuser to have that permission, so you can upload to OneDrive and share the link. 

[AllisonKennedy · Accepted solution]

@Believer 

You can filter the measure instead of the visual: 

Cust_RecID_Count = IF(ISBLANK([Customer Count]), BLANK(), CALCULATE([Customer Count], ALLEXCEPT('Table', 'Table'[Customer  ], 'Table'[RecID  ])) )

[Believer]

@AllisonKennedy ,

If I remove "Customer" (abbreviation) from the table, the counts are incorrect.  I understand this is because I removed "Customer" from the row context, so now I am seeing count of each RecID period.  How do I get Customer back into the row context without having it displayed?  (Given the relationship between the tables, I thought the context would have been impled.)  I could leave the column in and set the width to zero, but again that feels like a hack...

[Believer]

@AllisonKennedy 

Thank you very much, such a simple fix!  Pretty sure this is also the answer to another question I've had open for a while with no answers... 

I thought there may have been something more extravagant at work here.  I'm still trying to absorb the nuance of CALCULATE filters.  I recently saw that calculate filters are always tables (thanks SQLBI!) and thought somehow this was multiplying the table siginifcantly (given all the extra rows appearing) and that I was using the incorrect approach.