Drastic CTE query speed degradation on actions in outer select - sql

Question

I have a complex report (Similar to example below) it returns with only 52 items in 42 seconds (Slow but does complex joins) it drastically slows down when I do the following:
Add a column in the outer select to get a serialized list of items for
each item in the result (Adds 150 seconds) - See function.
Insert result into a variable table (Adds 120 seconds), I though hoped flattening it would reduce the (1) issue but no cigar.
How I understand the execution is that the where is executed first then logic in the select (Should only be done for the 52 result items). However if I run that exact 52 items with the (1) scenario it only takes 7 seconds vs the added 150 when using it on the outer CTE select.
Why could this be and how can I add the column without the bloated execution time?
CREATE FUNCTION PriorShippers
(
#customerId nchar(5)
)
RETURNS varchar(500)
AS
BEGIN
DECLARE #return varchar(500);
with data as
(
select distinct S.CompanyName from Customers C
join Orders O on C.CustomerID = O.CustomerID
join Shippers S on O.ShipVia = S.ShipperID
where C.CustomerID = #customerId
) select #return = STUFF((select CompanyName + ' ' from data FOR XML PATH('')),1,0,'')
return #return
END
The query (I used Northwind database - Install Instructions here for SQL 2012)
DECLARE #categories TABLE
(
Name varchar(100),
SourceCountry varchar(100)
);
insert into #categories VALUES ('Seafood', 'US');
insert into #categories VALUES ('Beverages', 'US');
insert into #categories VALUES ('Condiments', 'US');
insert into #categories VALUES ('Dairy Products', 'India');
insert into #categories VALUES ('Grains/Cereals', 'India');
with data as
(
select C.CustomerID, C.CompanyName,
(CASE WHEN EXISTS(select * from Orders O where O.CustomerID = C.CustomerID) THEN
(select count(distinct CAT.CategoryID) from Orders O
join [Order Details] OD on O.OrderID = OD.OrderID
join Products P on OD.ProductID = P.ProductID
join Categories CAT on P.CategoryID = CAT.CategoryID
where EXISTS(select * from #categories where Name = CAT.CategoryName AND SourceCountry = 'US'))
ELSE 0 END) as 'US Orders',
(CASE WHEN EXISTS(select * from Orders O where O.CustomerID = C.CustomerID) THEN
(select count(distinct CAT.CategoryID) from Orders O
join [Order Details] OD on O.OrderID = OD.OrderID
join Products P on OD.ProductID = P.ProductID
join Categories CAT on P.CategoryID = CAT.CategoryID
where EXISTS(select * from #categories where Name = CAT.CategoryName AND SourceCountry = 'India'))
ELSE 0 END) as 'India Orders'
from Customers C
) select top 10 CompanyName, [US Orders], [India Orders]
-- Below: Adding this have significant slow down
, dbo.PriorShippers(CustomerID)
from data where [US Orders] > 0 Order By [US Orders]

Answer 1

You are executing a user-defined function for each selected row. In short, don't do that.
Move the code in the function into a DTE directly within the report.
Also, I noticed that the joins for your counts are identical. They look like prime candidates for a third DTE that you could simply join to to further improve the performance of the overall query.