Active IDs in date ranges - oracle

Question

I have requirement to get number of customers active in a month based on their revenue contribution period.
Original Data:
ACCOUNT_ID REVENUE_START_DATE REVENUE_END_DATE
1234 1/14/2010 0:00 4/13/2010 23:59
4567 2/9/2010 0:00 3/8/2010 23:59
1234 5/9/2010 0:00 6/8/2010 23:59
Expected Result
Month Count
Dec-09 0
Jan-10 1
Feb-10 2
Mar-10 2
Apr-10 1
May-10 1
Jun-10 1
Jul-10 0
Aug-10 0
Sep-10
Oct-10
Below is the oracle code I worked on (with help of google) but I am not getting correct result due to overlapping dates. I request the experts to help me with this. (Thanks in Advance)
Current Result:
YEAR_ MONTH_ ACT
2010 January 2
2010 February 3
2010 March 3
2010 April 3
ORACLE CODE:
with tab as
(
select distinct ACCOUNT_ID, billing_start_date as revenue_start_date, billing_end_date as revenue_end_date
from accounts
),
year_tab as
(
select
add_months(min_date, level -1) m
from
(
select min(trunc(revenue_start_date,'YYYY')) min_date, add_months(max(trunc(revenue_end_date,'YYYY')), 12) max_date
from tab
)
connect by level <= months_between(max_date, min_date)
)
select to_char(m,'YYYY') year_,
to_char(m,'Month') month_,
nvl(act, 0) act
from year_tab,
(
select m date_,count(*) act
from tab, year_tab
where m between trunc(revenue_start_date,'MM') and trunc(revenue_end_date,'MM')
group by m
) month_tab
where m = date_(+)
order by m;

Answer 1

It's taken me a while to see why you think there is a problem. With the original three rows of data you supplied, running your query gives exactly your 'expected result'. With the 54 rows of data from your CSV file, the result is 48 rows (covering four years), with non-zero totals from January 2010 to January 2013. The first few rows returned are:
YEAR_ MONTH_ ACT
----- ------------------------------------ ----------
2010 January 2
2010 February 3
2010 March 3
2010 April 3
2010 May 2
But that looks correct:
select * from accounts
where not (billing_start_date > date '2010-02-01'
or billing_end_date < date '2010-01-01');
ACCOUNT_ID BILLING_START_DATE BILLING_END_DATE
---------- ------------------ ------------------
1234 09/01/2010 00:00 08/02/2010 23:59
4567 14/01/2010 00:00 13/04/2010 23:59
2 rows selected
select * from accounts
where not (billing_start_date > date '2010-03-01'
or billing_end_date < date '2010-02-01');
ACCOUNT_ID BILLING_START_DATE BILLING_END_DATE
---------- ------------------ ------------------
1234 09/01/2010 00:00 08/02/2010 23:59
4567 14/01/2010 00:00 13/04/2010 23:59
1234 09/02/2010 00:00 08/03/2010 23:59
3 rows selected
select * from accounts
where not (billing_start_date > date '2010-04-01'
or billing_end_date < date '2010-03-01');
ACCOUNT_ID BILLING_START_DATE BILLING_END_DATE
---------- ------------------ ------------------
4567 14/01/2010 00:00 13/04/2010 23:59
1234 09/02/2010 00:00 08/03/2010 23:59
1234 09/03/2010 00:00 08/04/2010 23:59
3 rows selected
But what I think you wanted wasn't really stressed in the question: 'to get number of customers active'. Assuming that by 'customer' you mean unique account IDs, you just need to modify the count:
select m date_,count(distinct account_id) act
from tab, year_tab
...
... which gives the first few rows as:
YEAR_ MONTH_ ACT
----- ------------------------------------ ----------
2010 January 2
2010 February 2
2010 March 2
2010 April 2
2010 May 1
What you were doing wrong was trying to apply the distinct in your tab subquery; but distinct returns distinct rows, and as the dates were different that wasn't actually reducing the number of rows returned.
Which doesn't quite match your expected result still, but does seem to match the data (if my assumption about what you want is right), and still does give your expected result for you three-row sample.
Another way to write the query, which I find a bit easier to follow, and using ANSI join syntax:
with t as (
select add_months(min_date, level - 1) month_start,
add_months(min_date, level) next_month_start
from (
select trunc(min(billing_start_date),'YYYY') min_date,
add_months(trunc(max(billing_start_date),'YYYY'), 12) max_date
from accounts
)
connect by level <= months_between(max_date, min_date)
)
select to_char(t.month_start,'YYYY') year_,
to_char(t.month_start,'Month') month_,
count(distinct a.account_id) act
from t
left join accounts a on not (billing_start_date > t.next_month_start
or billing_end_date < t.month_start)
group by t.month_start
order by t.month_start;