Grouping by the Hour and alerting based on count of usage per hour.

DMJerry

Hello,

 

I am trying to do a group by using hour of day.  The date operator in Magic does not have that funcationality but curious how I can do it within magic or in SQL.  Here is my data where I want to count how many times a code was use in an hour for each day I want to use.

COUPON-CODE	updated_at	Order #	Goal(Count of Usage by Hour)
10%	17-Apr-20 10:04:53	1	2
10%	17-Apr-20 10:43:49	2	2
10%	17-Apr-20 5:43:25	3	1
10%	17-Apr-20 0:46:53	4	1
10%	17-Apr-20 17:10:14	5	1
20%	17-Apr-20 2:51:38	6	1
20%	17-Apr-20 6:42:59	7	1
20%	17-Apr-20 13:24:49	8	2
20%	17-Apr-20 13:53:14	9	2
50%	17-Apr-20 15:51:46	10	1
50%	17-Apr-20 16:02:26	11	3
50%	17-Apr-20 16:27:03	12	3
50%	17-Apr-20 16:27:03	13	3

jaeW_at_Onyx

In Magic:  create a copy of the column converted to TEXT and do a REGEX / STRING transform.

In MySQL: use date_part functions (like hour).

 

Best Solution:  split date and time (HH:mm) into separate columns during data ingestion.  THEN use a Fusion to join a Time Dimension table (where you have one row for every minute of the day).

 

Why is Fusion the best model?  

Because it minimizes the number of times you have to read and write the data. AND by using a Time dimension you can create custom groupings like ("Morning", "coffee-break", "evening") without using a complex beast mode in each of your cards.

 

Why does splitting date_time make sense?

Because you'll want to see a trend in behavior on Saturdays versus trend in behavior on Mondays.  By using a Time dimension you can avoid ugly beast modes.

GrantSmith

Do you have the time stamp for the hours you're wanting to graph?  Domo now supports hour granularity in the time slicer so you could graph by hour. You could also possibly use the HOUR function in a beast mode might work as well

Unknown

Hello,

 

It just won't work for the alert.  You have to group it to give it a trigger.

GrantSmith

What are you attempting to alert on?

jaeW_at_Onyx

In Magic:  create a copy of the column converted to TEXT and do a REGEX / STRING transform.

In MySQL: use date_part functions (like hour).

 

Best Solution:  split date and time (HH:mm) into separate columns during data ingestion.  THEN use a Fusion to join a Time Dimension table (where you have one row for every minute of the day).

 

Why is Fusion the best model?  

Because it minimizes the number of times you have to read and write the data. AND by using a Time dimension you can create custom groupings like ("Morning", "coffee-break", "evening") without using a complex beast mode in each of your cards.

 

Why does splitting date_time make sense?

Because you'll want to see a trend in behavior on Saturdays versus trend in behavior on Mondays.  By using a Time dimension you can avoid ugly beast modes.

Unknown

Hi,

 

Trying to alert anytime I have a code used more than 3 times an hour.

Unknown

This is right but the issue now is that is goes from this to this:

2020-04-21T00:00:41

 

 

1:29

Apr 20, 2020 5:00:41 PM

 

Figuring that out right now but thanks for the solution!

jaeW_at_Onyx

@DMJerryVuori , 

I know this doesn't help you now, but given that you and several other users voiced challenges dealing with date and time data, i'm going to run a mini-hackathon on handling date and time in Domo next week thursday. 

 

You're welcome to join:  https://dojo.domo.com/t5/EMEA/DUG-Meeting-30-40-2020-1430-UTC-BUILD-IT-Working-with-DateTime/m-p/48274#M11