SQL 百分比 - 条件总和

Question

I want to calculate percentage churn and retain values according to app language. I am using Snowflake.

My current is as follows:

SELECT i.language, lb.churn_status, SUM(lb.total_users) AS user_count
FROM level_based_churn_retain_comparison AS lb
    INNER JOIN sessions AS s
    ON s.level_index=lb.level_index
        INNER JOIN installs AS i
        ON s.user_id=i.user_id

GROUP BY i.language, lb.churn_status
ORDER BY i.language, user_count

And the output is

language churn_status user_count

Answer 1

So with this very imaginative data (that I have in a CTE to make it easy for me)

with level_based_churn_retain_comparison(level_index, total_users, churn_status) as (
    select * from values
    (10, 100, 'a'),
    (20, 400, 'a'),
    (30, 300, 'a')
), installs(user_id, language) as (
    select * from values
    (1, 'en'),
    (2, 'en'),
    (3, 'fr')
), sessions(user_id, level_index) as (
    select * from values
    (1, 10),
    (2, 20),
    (3, 30)
)

your SQL + RATIO_TO_REPORT

SELECT 
    i.language, 
    lb.churn_status, 
    SUM(lb.total_users) AS user_count,
    100.0 * ratio_to_report(user_count) over() as percent_churn
FROM level_based_churn_retain_comparison AS lb
INNER JOIN sessions AS s
    ON s.level_index = lb.level_index
INNER JOIN installs AS i
    ON s.user_id = i.user_id
GROUP BY i.language, lb.churn_status
ORDER BY i.language, user_count

we get:

LANGUAGE	CHURN_STATUS	USER_COUNT	PERCENT_CHURN
en	a	500	62.5
fr	a	300	37.5

Now if you want the percentage to be per churn_status, we can put that into the PARTITION BY clause, like so:

with level_based_churn_retain_comparison(level_index, total_users, churn_status) as (
    select * from values
    (10, 100, 'a'),
    (20, 400, 'b'),
    (30, 300, 'a')
), installs(user_id, language) as (
    select * from values
    (1, 'en'),
    (2, 'en'),
    (3, 'fr')
), sessions(user_id, level_index) as (
    select * from values
    (1, 10),
    (2, 20),
    (3, 30)
)
SELECT 
    i.language, 
    lb.churn_status, 
    SUM(lb.total_users) AS user_count,
    100.0 * ratio_to_report(user_count) over(partition by lb.churn_status) as percent_churn
FROM level_based_churn_retain_comparison AS lb
INNER JOIN sessions AS s
    ON s.level_index = lb.level_index
INNER JOIN installs AS i
    ON s.user_id = i.user_id
GROUP BY i.language, lb.churn_status
ORDER BY i.language, user_count

giving:

LANGUAGE	CHURN_STATUS	USER_COUNT	PERCENT_CHURN
en	a	100	25
en	b	400	100
fr	a	300	75