SQL: Get date when cumulative sum reaches a mark - sql

Question

I have a table in the following format:
APP_iD| Date | Impressions
113 2015-01-01 10
113 2015-01-02 5
113 2015-01-03 50
113 2015-01-04 35
113 2015-01-05 30
113 2015-01-06 75
Now, I need to know the date when cumulative SUM of those impressions crossed 65/100/150 and so on.
I tried using CASE WHEN statement:
CASE WHEN SUM(impressions) >100
THEN date
but it doesn't sum the data across the column. It just does checks against the individual row.
My final result should look like:
APP_iD | Date_65 | Date_100 | Date_150
113 2015-01-03 2015-01-04 2015-01-06
Does anyone know how to do this?
Is this even possible?

Answer 1

Use sum() over() to get the running sum and check for the required values with a case expression. Finally aggregate the results to get one row per each app_id.
select app_id,max(dt_65),max(dt_100),max(dt_150)
from (
select app_id
,case when sum(impressions) over(partition by app_id order by dt) between 65 and 99 then dt end dt_65
,case when sum(impressions) over(partition by app_id order by dt) between 100 and 149 then dt end dt_100
,case when sum(impressions) over(partition by app_id order by dt) >= 150 then dt end dt_150
from t) x
group by app_id

Answer 2

with c as (
select
app_id, date,
sum(impressions) over (partition by app_id order by date) as c
from t
)
select app_id, s65.date, s100.date, s150.date
from
(
select distinct on (app_id) app_id, date
from c
where c >= 65 and c < 100
order by app_id, date
) s65
left join
(
select distinct on (app_id) app_id, date
from c
where c >= 100 and c <150
order by app_id, date
) s100 using (app_id)
left join
(
select distinct on (app_id) app_id, date
from c
where c >= 150
order by app_id, date
) s150 using (app_id)
;
app_id | date | date | date
--------+------------+------------+------------
113 | 2015-01-03 | 2015-01-04 | 2015-01-06
Without the pivot:
select distinct on (app_id, break) app_id, break, date
from (
select *,
case
when c < 100 then 65
when c < 150 then 100
else 150
end as break
from (
select
app_id, date,
sum(impressions) over (partition by app_id order by date) as c
from t
) t
where c >= 65
) t
order by app_id, break, date
;
app_id | break | date
--------+-------+------------
113 | 65 | 2015-01-03
113 | 100 | 2015-01-04
113 | 150 | 2015-01-06

Answer 3

You can try this for desired result.
with t as (select app_id, date, sum(Impressions)
over (partition by app_id order by date) AS s from tbl)
select app_id,
min(date_65) AS date_65 ,
min(date_100) AS date_100,
min(date_150) AS date_150
-- more columns to observe other sum of Impressions
from
(select app_id,
CASE WHEN (s >= 65 and s < 100) THEN date END AS date_65,
CASE WHEN (s >= 100 and s < 150) THEN date END AS date_100,
CASE WHEN (s >= 150 ) THEN date END AS date_150
-- more cases to observe other sum of Impressions
from t) q
group by q.app_id
if you want to observe more sum of Impressions, just add more conditions