Finding a minimum date before another date - sql

Question

Let's say I have two tables. One is a table with information about customer service inquiries, which contains information about the customer and the time the inquiry was placed. The customer's information (in this case, the ID) is saved for all future inquiries.
CUST_ID INQUIRY_ID INQUIRY_DATE
001 34 2015-05-03 08:15
001 36 2015-05-05 13:12
002 39 2015-05-10 18:43
003 42 2015-05-12 14:58
003 46 2015-05-14 07:27
001 50 2015-05-18 19:06
003 55 2015-05-20 11:40
The other table contains information about the resolution dates for all customer inquiries.
CUST_ID RESOLVED_DATE
001 2015-05-06 12:54
002 2015-05-11 08:09
003 2015-05-14 19:37
001 2015-05-19 16:12
003 2015-05-22 08:40
The resolution table doesn't have a key to link to the inquiry table other than the CUST_ID, so in order to calculate the time to resolution, I want to determine the minimum inquiry date before the resolution for EACH resolution date. The resulting table would look like this:
CUST_ID FIRST_INQUIRY RESOLVED_DT
001 2015-05-03 08:15 2015-05-06 12:54
001 2015-05-18 19:06 2015-05-19 16:12
002 2015-05-10 18:43 2015-05-11 08:09
003 2015-05-12 14:58 2015-05-14 19:37
003 2015-05-20 11:40 2015-05-22 08:40
At first I just went with min(case when INQUIRY_DATE < RESOLVED_DT), but for people like customers 001 and 003 who have multiple inquiries across different dates, the query would just return the first ever inquiry date, not the first since the last inquiry. Does anyone know how to do this? I'm using Netezza.

Answer 1

One option is to create a subquery for each table (inquries and resolutions) which numbers the transaction for each CUST_ID using the date. Then, the two subqueries can be joined together using this ordered index column along with the CUST_ID.
I also used the INQUIRY_ID in the inquiries table to break a tie, should it occur. There is not way to break a tie in the resolutions table for a given customer and date based on the data you showed us.
SELECT t1.CUST_ID, t1.INQUIRY_ID AS FIRST_INQUIRY, t2.RESOLVED_DATE AS RESOLVED_DT
FROM
(
SELECT CUST_ID, INQUIRY_ID, INQUIRY_DATE,
(SELECT COUNT(*) + 1
FROM inquiries
WHERE CUST_ID = t.CUST_ID AND INQUIRY_DATE <= t.INQUIRY_DATE
AND INQUIRY_ID < t.INQUIRY_ID) AS index
FROM inquiries AS t
) AS t1
INNER JOIN
(
SELECT CUST_ID, RESOLVED_DATE,
(SELECT COUNT(*) + 1
FROM resolutions
WHERE CUST_ID = t.CUST_ID AND RESOLVED_DATE < t.RESOLVED_DATE) AS index
FROM resolutions t
) AS t2
ON t1.CUST_ID = t2.CUST_ID AND t1.index = t2.index
Here are what the subquery tables look like:
inquiries:
CUST_ID INQUIRY_ID INQUIRY_DATE index
001 34 2015-05-03 08:15 1
001 36 2015-05-05 13:12 2
002 39 2015-05-10 18:43 1
003 42 2015-05-12 14:58 1
003 46 2015-05-14 07:27 2
001 50 2015-05-18 19:06 3
003 55 2015-05-20 11:40 3
resolutions:
CUST_ID RESOLVED_DATE index
001 2015-05-06 12:54 1
002 2015-05-11 08:09 1
003 2015-05-14 19:37 1
001 2015-05-19 16:12 2
003 2015-05-22 08:40 2
Note that this solution is not robust to missing data, e.g. there is an inquiry which was not closed, or the resolution was never recorded.