SQL Query to group ID overlap (via inner join) by month - sql

Question

I'm trying to find a query that will give me the number of customers that have transacted with 2 different entities in the same month. In other words, customer_ids that transacted with company_a and company_b within the same month. Here is what I have so far:
SELECT Extract(year FROM company_a_customers.transaction_date)
|| Extract(month FROM company_a_customers.transaction_date) AS
payment_month,
Count(UNIQUE(company_a_customers.customer_id))
FROM (SELECT *
FROM my_table
WHERE ( merchant_name LIKE '%company_a%' )) AS company_a_customers
INNER JOIN (SELECT *
FROM my_table
WHERE ( merchant_name = 'company_b' )) AS
company_b_customers
ON company_a_customers.customer_id =
company_b_customers.customer_id
GROUP BY Extract(year FROM company_a_customers.transaction_date)
|| Extract(month FROM company_a_customers.transaction_date)
The problem is that this is giving me a running total of all customers that transacted with company A on a month-by-month basis who also ever transacted with company B.
If I whittle it down to a specific month, it will obviously give me the correct overlap, because the query is only getting IDs for that month:
SELECT Extract(year FROM company_a_customers.transaction_date)
|| Extract(month FROM company_a_customers.transaction_date) AS
payment_month,
Count(UNIQUE(company_a_customers.customer_id))
FROM (SELECT *
FROM my_table
WHERE ( merchant_name LIKE '%company_a%' )
AND transaction_date >= '2017-06-01'
AND transaction_date <= '2017-06-30') AS company_a_customers
INNER JOIN (SELECT *
FROM my_table
WHERE ( merchant_name = 'company_b' )
AND transaction_date >= '2017-06-01'
AND transaction_date <= '2017-06-30') AS
company_b_customers
ON company_a_customers.customer_id =
company_b_customers.customer_id
GROUP BY Extract(year FROM company_a_customers.transaction_date)
|| Extract(month FROM company_a_customers.transaction_date)
How can I do this in one query to get monthly totals for customers who transacted with both companies within the given month?
Desired result: Output of second query, but for every month that is in the database. In other words:
January 2017: xx,xxx overlapping customers
February 2017: xx,xxx overlapping customers
March 2017: xx,xxx overlapping customers
Thanks very much.

Answer 1

You could simply calculate year/month for both and then add it as a join-condition, but this is not very efficient as it might create a huge intermediate result.
You better check for each month/customer if there were transactions with both merchants using conditional aggregation. And then count by month:
SELECT payment_month, count(*)
FROM
( SELECT Extract(year FROM transaction_date)
|| Extract(month FROM transaction_date) AS payment_month,
customer_id
FROM my_table
WHERE ( merchant_name LIKE '%company_a%' )
OR ( merchant_name = 'company_b' )
GROUP BY payment_month,
customer_id
-- both merchants within the same months
HAVING SUM(CASE WHEN merchant_name LIKE '%company_a%' THEN 1 ELSE 0 END) > 0
AND SUM(CASE WHEN merchant_name = 'company_b' THEN 1 ELSE 0 END) > 0
) AS dt
GROUP BY 1
YOur payment_month calculation is to complicated (and the returned string is not nicely formatted).
To get year/month as string:
TO_CHAR(transaction_date, 'YYYYMM')
as number:
EXTRACT(YEAR FROM transaction_date) * 100
+ EXTRACT(MONTH FROM transaction_date)
or calculate the first of month:
TRUNC(transaction_date, 'mon')

Answer 2

You should be able to get your desired results in one query just by counting the number of merchant_names per month per customer id. Using HAVING > 1 will show you only customers with transactions with both (or more if there are more matches for like '%company_a%').
SELECT
EXTRACT(Year from transaction_date)||EXTRACT(Month from transaction_date) as payment_month
,customer_id
,COUNT(DISTINCT merchant_name) as CompanyCount
FROM my_table
WHERE transaction_date >= '2017-06-01' AND transaction_date <= '2017-06-30'
AND (merchant_name = 'company_b' or merchant_name LIKE '%company_a%')
GROUP BY
EXTRACT(Year from transaction_date)||EXTRACT(Month from transaction_date)
,customer_id
HAVING COUNT(DISTINCT merchant_name) > 1