Insert Into with Select works in once instance, but not the other. Subquery error or incorrect data

Question

I am trying to write a script that will combine specific survey results based on the demographs used in the survey. I have been able to write a script that adds demographs together to create a new demograph e.g. M18+ and F18+ = A18+ which works perfectly however when trying to create a script that will create a demograph range e.g. A18-24, i am encountering some issues i cannot resolve.

My first script that adds two demographs together is as follows:

create table #SurveyTemp
(
 Com nchar(6),
 Survey nvarchar(50),
 Demo nchar(50),
 Wk int,
 Time int,
 Aud decimal(18,8)
);
insert into #SurveyTemp (Com, Survey, Demo, Wk, Time, Aud)
select Com, Survey, 'A18+', Wk, Time, sum(Aud)
from table_survey
where survey = 'LO2017'
and demograph like 'F18+'
or surveyid = 'LO2017'
and demograph like 'M18+'
group by Com, survey, Wk, Time:
insert into table_survey
select temp.Com, temp.Survey, temp.Demo, temp.Wk, temp.Time, temp.Aud
from #SurveyTemp temp
drop table #SurveyTemp

This script works fine. It adds two demographs together and inserts new rows of data, creating a sum of the audience for each Com, Survey, Wk, and then Time. See the sample data set below where the addition of M18+ and F18+ has created a new record called A18+

Let me start with a sample data set: (Disclaimer: the live database has MANY rows for each demograph)

com | surveyid | demo | wk | time | audience
-------------------------
 1 | LO2017 | A18+ | 1 | 300 | 4.7
 1 | LO2017 | F18+ | 1 | 300 | 1.9
 1 | LO2017 | M18+ | 1 | 300 | 2.8
 1 | LO2017 | A25+ | 1 | 300 | 2.3
 1 | LO2017 | A18+ | 2 | 100 | 3.7
 1 | LO2017 | F18+ | 2 | 100 | 1.9
 1 | LO2017 | M18+ | 2 | 100 | 2.8
 1 | LO2017 | A25+ | 2 | 100 | 4.3

What I need to write now is a script that creates a range. Notice how in the table above there is a A18+ and a A25+. Please bear in mind that in my live table there will be MULTIPLE rows for all demographs.

In this example, I will be using the above table. So, I need to create a similar script that will sum the audience for two demographs and then subtract one from the other to create a range. For example, subtract the A25+ data from A18+ to create a range called A1824. I hope that makes sense.

In the script below, I am trying to create a range called 'A1824' but I am either getting a "subquery returned more than 1 row" error OR it is inserting the wrong data into the temp table where the sum in AudOne and AudTwo is the SAME for every single Wk and Time.

create table #SurveyTemp
(
 Com nchar(6),
 Survey nvarchar(50),
 Demo nchar(50),
 Wk int,
 Time int,
 Aud decimal(18,8),
 AudOne decimal(18,8),
 AudTwo decimal(18,8)
);
insert into #SurveyTemp (Com, Survey, Demo, Wk, TimeBlock, Aud, AudOne, AudTwo)
(select Com, Survey, 'A1824', Wk, Time, 0.0, 
 (select sum(aud) from table_survey where demo = 'A18+'), 
 (select sum(aud) from table_survey where demo = 'A25+')
 from table_survey 
 where surveyid = 'LO2017'
 group by Com, survey, Wk, Time, 
)
update #SurveyTemp
set Aud = AudOne - AudTwo;
insert into table_survey
select temp.Com, temp.survey, temp.demo, temp.Wk, temp.Time, temp.aud
from #SurveyTemp temp
drop table #SurveyTemp

It is this second script that i need to get working and have spent hours trying to resolve without much luck. Please let me know if i can provide any more information to help you help me!

Edit: My expected results would be something like the following:

com | surveyid | demo | wk | time | audience
-------------------------
 1 | LO2017 | A1824 | 1 | 300 | 2.4

But there would be multiple rows where there are multiple 'wk' and 'time' entries. E.g.:

com | surveyid | demo | wk | time | audience
-------------------------
 1 | LO2017 | A1824 | 1 | 100 | 2.4
 1 | LO2017 | A1824 | 1 | 200 | 3.7
 1 | LO2017 | A1824 | 2 | 100 | 2.1
 1 | LO2017 | A1824 | 2 | 200 | 6.2

• Erik Reasonable Rates Darling: What are your expected results?
• L.Moy: @sp_BlitzErik Good question. I have added this into the question!
• McNets: As far as I understand time = sum(time) and wk is not necessary right?

Answer 1

I'll start after you has generated #SurveyTemp table.

I'm not sure to understand your expected result, take a look at your table ordered by demo:

SELECT * 
FROM @table_survey 
WHERE demo IN ('A18+','A25+') 
ORDER BY com, survey_id, demo, wk;
com | survey_id | demo | wk | time | audience | 
--: | :-------- | :--- | -: | ---: | :------- | 
 1 | LO2017 | A18+ | 1 | 300 | 4.7 | 
 1 | LO2017 | A18+ | 2 | 100 | 3.7 | 
 1 | LO2017 | A25+ | 1 | 300 | 2.3 | 
 1 | LO2017 | A25+ | 2 | 100 | 4.3 |

There are 2 records for each demo. If you want to add new record as a result of some operation between these two records, you should decide if you want to group by wk and time (that I think it's the correct way), or not.

In your insert command:

insert into #SurveyTemp (Com, Survey, Demo, Wk, TimeBlock, Aud, AudOne, AudTwo)
(select Com, Survey, 'A1824', Wk, Time, 0.0, 
 (select sum(aud) from table_survey where demo = 'A18+'), 
 (select sum(aud) from table_survey where demo = 'A25+')
 from table_survey 
 where surveyid = 'LO2017'
 group by Com, survey, Wk, Time, 
)

You're grouping by Com, Survey, Wk, Time, but you're calculating overall audience of all demo records, (without considering wk and time):

(select sum(aud) from table_survey where demo = 'A18+')
(select sum(aud) from table_survey where demo = 'A25+')

IMHO, at least wk should be added to the new records, and you should decide what to do with time.

My proposed solution: GROUP BY WK AND TIME

SELECT Com, survey_id, 'A1824', wk, time,
 (SELECT sum(audience) FROM @table_survey 
 WHERE demo = 'A18+' AND wk = t1.wk AND time = t1.time)
 - 
 (SELECT sum(audience) FROM @table_survey 
 WHERE demo = 'A25+' AND wk = t1.wk AND time = t1.time) AS audience
FROM @table_survey t1
WHERE survey_id = 'LO2017'
GROUP BY Com, survey_id, Wk, time;

NOTE: I've added wk and time to calculate sum(audience) for every demo.

Com | survey_id | (No column name) | wk | time | audience
--: | :-------- | :--------------- | -: | ---: | :-------
 1 | LO2017 | A1824 | 1 | 300 | 2.4 
 1 | LO2017 | A1824 | 2 | 100 | -0.6 

As you can see, first record match your expected result.

Without group by wk and time

SELECT Com, survey_id, 'A1824', 0 AS wk,
 avg(time) as time,
 (SELECT sum(audience) FROM @table_survey WHERE demo = 'A18+')
 - 
 (SELECT sum(audience) FROM @table_survey WHERE demo = 'A25+') AS audience
FROM @table_survey t1
WHERE survey_id = 'LO2017'
GROUP BY Com, survey_id;

This is the result:

Com | survey_id | (No column name) | wk | time | audience
--: | :-------- | :--------------- | -: | ---: | :-------
 1 | LO2017 | A1824 | 0 | 200 | 1.8 

NOTE: You don't need to INSERT & UPDATE, it can be done on a single INSERT operation.

Check it: dbfiddle here

• L.Moy: I... wow. That got it! Thank you so much for your well thought out and detailed response! Can i paypal you a beer?! haha.
• McNets: A Tequila shot will be enough!!