Postgresql query to return number of customers that made more than 1 order in a month - sql

Question

so I have this query in my code:
SELECT month, total_orders, first_orders, first_orders::numeric / total_orders::numeric * 100 AS ratio
FROM (
SELECT month, COUNT(1) AS total_orders
FROM (
SELECT DISTINCT date_trunc('month', date) AS month, email
FROM shop_orders
WHERE status_code = 'CONFIRMED'
) AS customers
GROUP BY month
) AS total_orders_my_month
LEFT JOIN (
SELECT month, COUNT(1) AS first_orders FROM (
SELECT email, date_trunc('month', MIN(date)) AS month
FROM shop_orders
WHERE status_code = 'CONFIRMED'
GROUP BY email
) AS first_orders_by_month
GROUP BY month
) AS first_orders_by_month2 USING (month)
ORDER BY month DESC
What it does is it retursn rows in a form: month | number of customers that month | number of first time customers that month | percentage of first time customers
What I need is find for each month also number of customers that made more then 1 order.
So I added this code:
LEFT JOIN (
SELECT month, COUNT(1) AS multiple_orders FROM (
SELECT email, date_trunc('month', MIN(date)) AS month
FROM shop_orders
WHERE status_code = 'CONFIRMED'
GROUP BY email HAVING COUNT(1) >=2
) AS multiple_orders_by_month
GROUP BY month
) AS multiple_orders_by_month2 USING (month)
This however returns only customers with 2 or more orders out of first-time customers not all customers. Can someone help please?
NOTES: All data is in shop_orders table, unique customer is identified by email field, date field has info about date of order.
I hope it's clear.
I am pretty new to postgresql so can someone point to the right direction? Thanks.

Answer 1

I would first like to point out that I am looking at this from a purely SQL route rather than postsgresql but the theory is pretty much the same. Your sub query SQL is probably overly complicated, here is my working on how I would create the query:
First, get a list of all e-mail addresses and how many orders each made that month; the following query will return the number of orders assigned to each e-mail:
SELECT
month
, email
, COUNT(*)
FROM
shop_orders
WHERE
status_code = 'Confirmed'
GROUP BY
month
, email
This will return for each month the order count for each e-mail address. You can then limit this list to all e-mails which have more than one per month using the HAVING filter method.
SELECT
month
, email
, COUNT(*) as ordersThisMonth
FROM
shop_orders
WHERE
status_code = 'Confirmed'
GROUP BY
month
, email
HAVING
COUNT(*) > 1
This sub-query could then be accessed using either another aggregate such as SUM(CASE WHEN ordersThisMonth > 1 THEN 1 ELSE 0 END) which will only count when more than one order is received. Personally I would try simplifying your query into two sub queries, one containing details which are specific to a month, the other for the month AND email.
(The following code has not been syntax checked and as previously mentioned I am looking at this from a purely SQL point of view.)
SELECT
month
, total_orders
, first_orders
/* Add together all firstMonthCount entries as it should only be 1 for the customers first month */
, SUM(firstMonthCount)::numeric / total_orders::numeric * 100 AS ratio
/* Count only e-mails which have more than one order for the month */
, SUM(CASE WHEN orderCount > 1 THEN 1 ELSE 0 END) AS multipleOrders
FROM
/*The following sub-query gets all the aggregates specific to a month */
( SELECT
date_trunc('month', date) AS month
, COUNT(1) AS total_orders
FROM
shop_orders
WHERE
status_code = 'CONFIRMED'
GROUP BY
month
) AS month_attributes
/*The following sub-query gets all the aggregates specific to a month AND e-mail */
LEFT JOIN (
SELECT
date_trunc('month', date) AS month
, email
, CASE WHEN date_trunc('month', date) = date_trunc('month', MIN(date)) THEN 1 ELSE 0 END AS firstMonthCount
, COUNT(*) AS orderCount
FROM
shop_orders
WHERE
status_code = 'CONFIRMED'
GROUP BY
month
,email
) AS email_month_attributes USING (month)
GROUP BY
month
Hopefully the above query and notes regarding the HAVING will help you find a solution that suits your purposes. I'm sure the above query could be simplified further if necessary for readability/performance.