How to use case statement with group by? - sqlite

Question

Assume our company has multiple marketing campaigns for one specific product, which might boost the sales of it in certain way. The results of the campaign are shown in the following data table:
Date CampaignID QtySold
2017-01-05 1 20
2017-01-18 2 35
2017-01-23 1 15
…
For modeling purposes, the desired output table looks like this:
CampaignID JanQtySold FebQtySold … DecQtySold
1 55 30
2 45 20
…
N
I have tried to get the month of each transaction for each campaign, and then group by campaignID and month.
select
campaignid,
strftime('%m',date) as Month,
sum(qtysold) as Sum_Qty
from campaign
group by campaignid, month
;
The return should include a unique CampaignID but I don't know how to proceed.

Answer 1

I believe that you want something like :-
select
campaignid,
coalesce(
(
SELECT sum(qtysold) FROM campaign AS c2 WHERE campaign.campaignid = c2.campaignid AND CAST(strftime("%m",date) AS INTEGER) = 1
)
,0) AS JanQtySold,
coalesce(
(
SELECT sum(qtysold) FROM campaign AS c2 WHERE campaign.campaignid = c2.campaignid AND CAST(strftime("%m",date) AS INTEGER) = 2
)
,0) As FebQtySold,
coalesce(
(
SELECT sum(qtysold) FROM campaign AS c2 WHERE campaign.campaignid = c2.campaignid AND CAST(strftime("%m",date) AS INTEGER) = 3
)
,0) As MarQtySold
/* .............. and so on ..........*/
from campaign
group by campaignid
;
That is introducing more groups is not really what you want for more derived columns (introducing more grouping components will introduce more rows, so you only want the 1 grouping component i.e. the campaign id). Rather you want to use sub-queries to generate the data for the derived columns.
Note for brevity only 3 months (Jan-Mar) have been shown, the other months is just a matter of copying one of the months and then amending the test value and the column name respectively.
Example :-
Using :-
DROP TABLE IF EXISTS campaign;
CREATE TABLE IF NOT EXISTS campaign (Date TEXT, CampaignID INTEGER, QtySold INTEGER);
INSERT INTO campaign VALUES
('2017-01-05',1,20),('2017-01-23',1,15),('2017-02-01',1,5),
('2017-01-18',2,35)
;
select
campaignid,
coalesce(
(
SELECT sum(qtysold) FROM campaign AS c2 WHERE campaign.campaignid = c2.campaignid AND CAST(strftime("%m",date) AS INTEGER) = 1
)
,0) AS JanQtySold,
coalesce(
(
SELECT sum(qtysold) FROM campaign AS c2 WHERE campaign.campaignid = c2.campaignid AND CAST(strftime("%m",date) AS INTEGER) = 2
)
,0) As FebQtySold,
coalesce(
(
SELECT sum(qtysold) FROM campaign AS c2 WHERE campaign.campaignid = c2.campaignid AND CAST(strftime("%m",date) AS INTEGER) = 3
)
,0) As MarQtySold
from campaign
group by campaignid
;
results in :-

Answer 2

You need conditional aggregation with a CASE statement inside SUM():
select
CampaignID,
sum(case strftime('%m',Date) when '01' then QtySold else 0 end) as JanQtySold,
sum(case strftime('%m',Date) when '02' then QtySold else 0 end) as FebQtySold,
sum(case strftime('%m',Date) when '03' then QtySold else 0 end) as MarQtySold,
sum(case strftime('%m',Date) when '04' then QtySold else 0 end) as AprQtySold,
sum(case strftime('%m',Date) when '05' then QtySold else 0 end) as MayQtySold,
sum(case strftime('%m',Date) when '06' then QtySold else 0 end) as JunQtySold,
sum(case strftime('%m',Date) when '07' then QtySold else 0 end) as JulQtySold,
sum(case strftime('%m',Date) when '08' then QtySold else 0 end) as AugQtySold,
sum(case strftime('%m',Date) when '09' then QtySold else 0 end) as SepQtySold,
sum(case strftime('%m',Date) when '10' then QtySold else 0 end) as OctQtySold,
sum(case strftime('%m',Date) when '11' then QtySold else 0 end) as NovQtySold,
sum(case strftime('%m',Date) when '12' then QtySold else 0 end) as DecQtySold
from campaign
where strftime('%Y',Date) = '2017'
group by CampaignID
I added the condition:
where strftime('%Y',Date) = '2017'
just in case there are rows in the table for multiple years.
If you don't want zeros in the results just remove all else 0 from the case statements.
See the demo.
Just to make it more efficient, you can a use a CTE:
with cte as (
select
CampaignID,
strftime('%Y',Date) as year,
strftime('%m',Date) as month,
QtySold
from campaign
)
select
CampaignID,
sum(case month when '01' then QtySold else 0 end) as JanQtySold,
sum(case month when '02' then QtySold else 0 end) as FebQtySold,
sum(case month when '03' then QtySold else 0 end) as MarQtySold,
sum(case month when '04' then QtySold else 0 end) as AprQtySold,
sum(case month when '05' then QtySold else 0 end) as MayQtySold,
sum(case month when '06' then QtySold else 0 end) as JunQtySold,
sum(case month when '07' then QtySold else 0 end) as JulQtySold,
sum(case month when '08' then QtySold else 0 end) as AugQtySold,
sum(case month when '09' then QtySold else 0 end) as SepQtySold,
sum(case month when '10' then QtySold else 0 end) as OctQtySold,
sum(case month when '11' then QtySold else 0 end) as NovQtySold,
sum(case month when '12' then QtySold else 0 end) as DecQtySold
from cte
where year = '2017'
group by CampaignID
See the demo.
Results:
| CampaignID | JanQtySold | FebQtySold | MarQtySold | AprQtySold | MayQtySold | JunQtySold | JulQtySold | AugQtySold | SepQtySold | OctQtySold | NovQtySold | DecQtySold |
| ---------- | ---------- | ---------- | ---------- | ---------- | ---------- | ---------- | ---------- | ---------- | ---------- | ---------- | ---------- | ---------- |
| 1 | 35 | 65 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| 2 | 35 | 75 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |