SQL Sample Function Seems to be returning more than the specific percentage

Question

I am trying to assign a control status of true to 10% of distinct groups in my table. Everything runs fine, but when I view the results, the percentages are way off. I'm getting more like 1/3 of records with the true control status, instead of the desired 10%. I've been using sample based on the distinct group id, but please let me know if something looks off here. Thanks!

PS: This is SQL Snowflake

Code and results listed below :

update table 
    set control_status=true
    where group_id in
       (select DISTINCT(group_id) from table sample(10));

update table 
    set control_status=false
    where control_status is null;


select control_status, count(distinct(group_id)),count(distinct(person_id)), count(control_status) from table group by control_status;

Results

control_status	count(distinct(group_id))	count(distinct(person_id))
True	50,000	100,000
False	100,000	200,000

Answer 1

I think what you're looking for is the following - you need to create a representative set of the distinct group_ids in order to chose 10% of those and not be skewed by the rows in the underlying table that pertain to people who have a 1-M relationship with group_id:

update table_name 
    set control_status=true
    where group_id in
       (select group_id from (select distinct (group_id) from table_name) x sample(10));

The rest remains the same as your original post:

update table 
    set control_status=false
    where control_status is null;


select control_status, count(distinct(group_id)),count(distinct(person_id)), count(control_status) from table group by control_status;