Conditional COUNT within CASE statement - sql

Question

Table 1, is a list of clients, what membership they have, what service they used, and the date the service was used
Table 2, is just table 1 grouped by month and membership type, then a count of the service sessions
What I am trying to do is count membership sessions only by particular service types. This is what I have so far, it returns an error saying 'Service_Type' is not in an aggregate function or group by clause, when I put 'Service_Type' in a group by, the query has no errors but the SESSIONS column is all NULL.
SELECT
DATEFROMPARTS(YEAR(t1.Date),MONTH(t1.Date),1)AS 'Draft_Date',
Membership,
CASE
WHEN Membership = 5 AND Service_Type = 'A' THEN COUNT(*)
WHEN Membership = 2 AND Service_Type IN ('J','C')
END AS'SESSIONS'
FROM Table1 t1
GROUP BY DATEFROMPARTS(YEAR(t1.Date),MONTH(t1.Date),1),Membership
The case statement will include all memberships and service types but I think this is enough for my example. Any help would be greatly appreciated! I've been on this for days.
Table 1
Table 2

Answer 1

You were nearly there! I've made a few changes:
SELECT
DATEFROMPARTS(YEAR(t1.Date), MONTH(t1.Date),1) AS Draft_Date,
Membership,
COUNT(CASE WHEN t1.Membership = 5 AND t1.Service_Type = 'A' THEN 1 END) as m5stA,
COUNT(CASE WHEN t1.Membership = 2 AND t1.Service_Type IN ('J','C') THEN 1 END) as m2stJC
FROM Table1 t1
GROUP BY YEAR(t1.Date), MONTH(t1.Date), Membership
Changes:
Avoid using apostrophes to alias column names, use ascii standard " double quotes if you must
When doing a conditional count, put the count outside the CASE WHEN, and have the case when return something (any non null thing will be fine - i used 1, but it could also have been 'x' etc) when the condition is met. Don't put an ELSE - CASE WHEN will return null if there is no ELSE and the condition is not met, and nulls don't COUNT (you could also write ELSE NULL, though it's redundant)
Qualify all your column names, always - this helps keep the query working when more tables are added in future, or even if new columns with the same names are added to existing tables
You forgot a THEN in the second WHEN
You don't necessarily need to GROUP BY the output of DATEFROMPARTS. When a deterministic function is used (always produces the same output from the same inputs) the db is smart enough to know that grouping on the inputs is also fine
Your example data didn't contain any data that would make the COUNT count 1+ by the way, but I'm sure you will have other conditional counts that work out (it just made it harder to test)

Answer 2

use sum
SELECT DATEFROMPARTS(YEAR(t1.Date),MONTH(t1.Date),1) AS Draft_Date , Membership,
sum(CASE WHEN Membership = 5 AND Service_Type = 'A' THEN 1 else 0 end),
sum(case WHEN Membership = 2 AND Service_Type IN ('J','C') then 1 else 0 end)
FROM Table1 t1 group by DATEFROMPARTS(YEAR(t1.Date),MONTH(t1.Date),1)