Forum Discussion

Atseaukes's avatar
Atseaukes
Frequent Visitor
8 years ago
Solved

Index column based on date, multiple entries excluded based on time

Current colomns        Desired column action id user_id date time   Index 2ac9756d-d46d-42ff-b759-6fb88beaf8f7 1 11-06-18 11:35:40   1 30ac92ed-8e81-4443-aed7-8261eb0a597c ...
  • Atseaukes's avatar
    Atseaukes
    8 years ago

    Step 1:

    Index =

    VAR UserFirstDate = MINX(FILTER('Events Completion';'Events Completion'[user_id] = EARLIER('Events Completion'[user_id]));'Events Completion'[DateTime])
    RETURN

    DIVIDE (DATEDIFF(UserFirstDate ;'Events Completion'[DateTime];DAY);1)+1

     

    Step 2:

    Rank =
    VAR d = 'Events Completion'[Index]
    VAR c = 'Events Completion'[user_id]
    RETURN
    CALCULATE (
    RANK.EQ ( d; 'Events Completion'[Index]; ASC );
    FILTER ( ALL ( 'Events Completion'); 'Events Completion'[user_id] = c ))

Phil_Seamark's avatar
Phil_Seamark
Icon for Microsoft Employee rankMicrosoft Employee
8 years ago

Hi Atseaukes

 

Is the time of day important?

 

Eg, if the first item starts at 2pm (in the afternoon) on a Monday.

 

In that case, what value would you give to an event that took place at 10am on the following Wednesday, and also to one that took place at 8pm on the same Wednesday?  

 

Do you need to track that one of those events was less than 48 hours while the other is more than 48 hours to give a different result?  Or should both have the same value because the time component is not important?

 

 

Atseaukes's avatar
Atseaukes
Frequent Visitor
8 years ago

Hi Phil_Seamark,

 

The time of day is not important.

 

Monday would have value 1. The following Wednesday at 10am would have value 2, just like the event at 8pm (value = 2).

  • Phil_Seamark's avatar
    Phil_Seamark
    Icon for Microsoft Employee rankMicrosoft Employee
    8 years ago

    ok, then does this get closer?

     

    Column = 
VAR UserFirstDate = MINX(FILTER('Table1','Table1'[user_id] = EARLIER('Table1'[user_id])),'Table1'[date] )
RETURN DATEDIFF(UserFirstDate ,'Table1'[date] ,DAY)+1
    • Atseaukes's avatar
      Atseaukes
      Frequent Visitor
      8 years ago

      Hi Phil_Seamark,

       

      On the subsequent days the value is correct. However, if there are more days in between, those days are added to the value. The values must be an orderly one. 

       

      Like in this example (all have the same user_id):

      1-06-18, 12:00 ; value 1

      1-06-18, 13:00 ; value 1

      2-06-18, 12:00 ; value 2

      4-06-18, 12:00 ; value 3 (instead of value 4, which will be the result of the formula I posted)

      • Atseaukes's avatar
        Atseaukes
        Frequent Visitor
        8 years ago

        Hi Phil_Seamark,

         

        Do you have any suggestions what to do else?

        Maybe doable with a RANKX? 

        Rank = RANKX('Events Completion';'Events Completion'[Index];;ASC;Dense)

        But first it needs to be filtered by 'Events Completion'[User_ID] I guess, right?