MySQL: Aggregate weekly statistics with substatics (subqueries) - mysql

Question

I would like to gather weekly statics on a MySQL-Table.
The table itself has the following structure:
user_id action_id created
0 123 2017-01-01 00.00:00
0 124 ...
1 123 ...
... ... ...
I would like to aggregate the weekly statics for:
How many user where active per week
This is rather simple:
SELECT
YEARWEEK(created) as week,
COUNT(DISTINCT user_id) AS count
FROM data
GROUP BY YEARWEEK(created);
Additionally I could apply a sorting.
The result looks like:
week count
201701 2
201702 3
How many user where active per week for the very first time
I thought about solving it by using a subquery
SELECT
YEARWEEK(created) as week,
COUNT(DISTINCT user_id) AS count,
(
SELECT
COUNT(DISTINCT d2.user_id)
FROM data d2
WHERE YEARWEEK(d2.created) = week
AND NOT EXISTS (SELECT 1 FROM data d3
WHERE YEARWEEK(d3.created) < week AND d2.user_id = d3.user_id)
) as countNewUsers
FROM data d1
GROUP BY YEARWEEK(created);
How many junior user where active per week
Junior users were active between 1 and 10 times before the related week
Similar to the one above, but with other subquery
How many power user where active per week
Senior users were active more than 10 times before the related week
This works as expected, but has a rather poor performance, since the subquery is evaluated before the grouping happens. With millions of rows in a table, this takes ages.
Does anybody have a better solution for this query, ideally returning all values in single result set?

Answer 1

I think all of your queries could derive from one 'intermediate' table. It would contain (yearweek, userid, count).
Users active per week: Pretty much the same query, but faster from this table.
Active for first time: Self-join ON userid and desired week versus MIN(yearweek)
Uses before the target week: ... SUM(count) WHERE ... < week GROUP BY userid
Use the above to determine which userids of Junior/Power.