Optimize Z Scores across multiple dimensions

So I have progressed and have managed to get the above parts working well and responsive. Example code at end of post if anyone is interested.

However, my question around how the Switch function processes would be amazing. 

I have two pickers 1) Numerator & 2) Denominator. These allow you to select what measure you want in the calculations for ZScores (or even could be a simple ratio). 

The code I am using to find the appropriate measure to use takes the [Numerator Type] directly from the Numerator Picker. I currently have circa 10 measures so far, however I would prefer to scale that up more. As soon as I replace the [No Products] from the refactored original code the speed goes from 400ms to 5000+ms.

Key Metrics Numerator = IF(HASONEVALUE('Key Metrics Numerator'[Numerator Type]),
 SWITCH(MAX('Key Metrics Numerator'[Numerator Type]), 

"Contacts", [No Contact],
"Products", [No Products],
"PDDs", [No PDD],
"Waste", [No Wasted],
"Refunded", [No Refunded],
"Income", [Income]

)) 

Hence this makes me question how the SWITCH function operates. 

If "PDDs" is selected in the picker does the SWITCH statement follow the below steps?

1. Calculate the Switching value (MATCHER)
2. Check if the MATCHER matches "Contacts"
3. Calculate the value of [No Contacts]
4. Check if the MATCHER matches "Products"
5. Calculate the value of [No Products]
6. Check if the MATCHER matches "PDDs"
7. Calculate & Use the value of [No Products]

Now I hope the above steps do not occur, I would hope that steps 3 & 5 doesn't need to occur. However my query get so much slower as soon as this switch statement is added. I do reference the [Key Metrics Numerator] a substaintial amount of times throughout my ZScore calculations due to the quantity of Dimensions used.

Is there another approach the works better than using SWITCH, or restructing the Switch statement above better?

Or a better way to reference the Numerator & Denominator fewer times in the below code.

REFACTORED CODE SOLUTION

Note that [No Contact] & [No Products] from the orginal post have been replaced with [Key Metrics Numerator] & [Key Metrics Denominator] to demonstrate the above problem.

This is my approach to building the base tables to calculate from (takes around 200 ms)

ZScore Refactored = 
VAR OutcomeYearSimpleTable =
	ADDCOLUMNS(ADDCOLUMNS(
    SUMMARIZE ('Dates Outcome',
        'Dates Outcome'[Financial Year]), 
        "Category", "Outcome Year"),
        "Encoded", TRIM(CONCATENATE([Category], CONCATENATE(" ", [Financial Year]))), 
        "Numerator" , [Key Metrics Numerator],
		"Denominator", [Key Metrics Denominator]
        )	
VAR OutcomeYearTable = 
	ADDCOLUMNS(ADDCOLUMNS(OutcomeYearSimpleTable,
		"NumeratorTotal", CALCULATE([Key Metrics Numerator], ALL('Dates Outcome'[Financial Year])),
	    "DenominatorTotal", CALCULATE([Key Metrics Denominator], ALL('Dates Outcome'[Financial Year]))
	),
		"IndexOfProportion", DIVIDE ( [Numerator], [NumeratorTotal]) - DIVIDE([Denominator], [DenominatorTotal])
	)	
VAR OutcomeYearTableZ = 
ADDCOLUMNS(
	ADDCOLUMNS(OutcomeYearTable,
		"Mean", AVERAGEX ( OutcomeYearTable, [IndexOfProportion] ),
		"StdDev", STDEVX.P ( OutcomeYearTable, [IndexOfProportion] )
	
	),
		"ZScore", DIVIDE (( [IndexOfProportion] - [Mean] ), [StdDev]) 
	)

Then I find this approach work well to pick which dimension to display (now from whatever column you want)

MINX(megaTable, if([Encoded] = MIN('Key Metrics Detail'[DimensionEncoded]), [ZScore], blank()))