As Ashleigh already noted, the biggest problem is that you have null values in `Occurred date`. Along those same lines, the larger dataset issue is that you are using a single column for double-duty, identifying the date as well as the activity simultaneously.

For example, in USERID = 2, there is no way to tell that the inactivity occurred in March without a datetime stamp associated with it. You need to restructure your dataset to create a denominator to divide by.

The most comprehensive solution would be to separate those two datapoints, and I would go with MarkSnodgrass's earlier suggestion and align your data to the Domo Calendar dataset. The SQL transform below should give you a starting point on how to recreate this dataset to get the calculations you want. I'm writing this code freehand, so forgive me if there are minior syntax or spelling errors. ?

----------

Table Transform 1: tmp_User

/* Create a unique list of all users */
SELECT DISTINCT `USERID`, `user group`
FROM tbl_UserActivity

----------

Table Transform 2: tmp_UserByDay

/* Create date record for every unique user */
SELECT
Cal.`dt`,
User.`USERID`,
User.`user group`
FROM tmp_User AS User
LEFT OUTER JOIN `raw_etumos_domo_calendar` AS Cal
ON Cal.`dt` >= '2020.01.01' -- Set this to how far in the past you want to report
AND `dt` < '2020.03.01' -- Set this to how far in the future you want to report

----------

Table Transform 3: tmp_UserActivityByDay

/* Populate activity records for every unique user */
SELECT
User.`dt`,
User.`USERID`,
User.`user group`,
(CASE WHEN Activity.`USERID` IS NOT NULL THEN 1 ELSE 0 END) AS Activity
FROM tmp_UserByDay AS User
LEFT OUTER JOIN tbl_UserActivity AS Activity
ON User.`USERID` = Activity.`USERID`
AND User.`dt` = Activity.`Occurred date`

----------

Output the above to a new dataset and you should be able to use a Beastmode calculation (example below) in your cards to get the percentages you are looking for. Let me know how it works for you.

Activity Rate = SUM(`Activity`) / COUNT(`USERID`)