DATEDIFF month interval not working as expected.

[PowerPaddy]

Hello again,

I'm using the DateDiff function to work out after how many months a policy was cancelled but it's not giving me the expected results.

Cancellation Month = DATEDIFF([POLICYSTARTDATE], [CANCELDATE], MONTH)

I'm expecting a 0 for policies that were cancelled within one month, ie where the difference between Date1 and Date2 is less than 1 month.

DAX seems to only return a 0 if Date1 and Date2 are in the same month, ie if Date1 is 31 Jan 17 and Date2 is 1 Feb 17 then the above formula returns 1. My understanding of this function is that it shouldn't work like this.

Here's an extract...

[Greg_Deckler]

Your understanding is incorrect. DATEDIFF counts the "interval boundaries" crossed between the two dates. Therefore, going from Jan 31st to Feb 1st crosses the month interval boundary (new month).

[PowerPaddy]

Thank you.

IBM gives the description of the DATEDIFF function on their website...
"The result will be the integer portion of the number of months between the two dates, with any fractional component truncated."

I will have to count number the number of days instead as a workaround.

[v-sihou-msft · Accepted solution]

@PowerPaddy

To get more accurate "MONTHDIFF", you can create a column like below:

Cancellation Month =
 ( 1
    * ( [CANCELDATE] - [POLICYSTARTDATE] ) )
    / DAY ( EOMONTH ( [CANCELDATE], 0 ) )

Regards,

[Greg_Deckler]

IBM is not the authority on DAX:

This is the authority:

https://msdn.microsoft.com/en-us/library/dn802538.aspx

[Harry_Long2]

How would you get it so that the function returns 0 if the difference between the two dates is less than one exact month? For example, if the start date is Jan 31st and the cancelled date is 1st Feb, we get 0 rather than 1. 

Thanks!

[Greg_Deckler]

Well, that's sort of going to depend on your definition of 1 exact month I suppose. You could probably get close with something like this:

Column = MROUND([DateCancelled]-[Date]*1.,30)/30