i have tested a experiment and i got Credible Interval: [-37.2%, 123%] , do you think its some problem in the setup?

its likely you have a small sample size

SELECT
2c.id AS value,
3c.created_at AS timestamp,
4c.employee_account_id AS user_id,
5c.company_id as company_id
6FROM
7  connection c
8  INNER JOIN account AS a ON a.id = c.employee_account_id
9 WHERE 
10 c.initiated_by = 'employee'

Here is the sql , the value is a id and i want to see what test generetes most ids

hi Samuel, sorry for the delay. yes, that should be okay

Okey then i dont need to have the COUNT? '

Hi Samuel, that's correct, you can omit the ​`COUNT()`​ if you use ​`1 as value`​ Below is an example SQL query for a Metric that uses the Count type:

SELECT
    user­_id,
    timestamp,
    1 as value
FROM
    purchases;

This SQL query is designed to count the number of purchases per user. It selects the ​`user­_id`​ and ​`timestamp`​ for each purchase event and assigns a constant value of ​`1`​ to each event. When this query is used in GrowthBook, the ​`value`​ column will be summed up to count the total number of purchase events per user.

Thanks!