Last Month calculations do not work in January

[jengwt]

I have a lot of calculation in a report which calculate stats for the "last [period of time]". Now that we have started a new year, a lot of these calculations seem broken.
I thought that these calcs would work in the new year because Dax's date/calendar math appears capable of handling math that spans different years and months. For instance, if you ask it for a date - 1 month and it the input as january, it correctly returns December. But this does not appear to be the case here.

Below you see a table detailing the numbers for certain periods of time. X1 is the average size of the population, X2 is the size of the losses in that period of time, and Ann Rate is the annualized rate of loss in that period. The periods in question are for the selected period (PRD) via a slicer, Month to Date, Year to Date, Last Month, Last Year, End of Month, and End of Year.

 

Here is the code for x1, the population of a given data set:

Population_LM = -- STATIC CALC
VAR LMDs = [LM_Days] -- Correctly retruns the number of days in the last month
VAR DN = [DNOW] -- DNOW being the current date
RETURN ROUND(CALCULATE(SUM('Population[COUNT])
, CALCULATETABLE('Population'
, MONTH('BI-DATE_DIM'[DIM_DATE]) = (MONTH(DN) - 1)
&& YEAR('BI-DATE_DIM'[DIM_DATE]) = YEAR(DN)
)
) / LMDs, 0)

x2, the count of losses:

Loss_LM = -- STATIC CALC
VAR DN = [DNOW]
RETURN CALCULATE(COUNT('Losses'[ID])
, CALCULATETABLE('Losses'
, (MONTH('BI-DATE_DIM'[DIM_DATE]) = MONTH(DN) - 1)
&& (YEAR('BI-DATE_DIM'[DIM_DATE]) = YEAR(DN))
)
)

 And Ann Rate is the annualized rate of loss, based on the calendar year:

TO_LM_ANN = ([Loss_LM]/[Population_LM]) * ([EOYrDay_LM] / [LM_Days]) -- STATIC calc. Works for all other metrics but this one.
-- [EOYrDay_LM] correcly returns the number of days for the year of which last month was a part.

In other words, the rate for the given period of time * the number of days in the year / the length of the period is the annualized rate. This calculation works for all other metrics, so the problem must be that the code doesn't know how to calculate stats for last month when the last month was in another year.

TABLES

+Sample data for @v-yuezhe-msft

Population is a table which for simplicity's sake has counts contributing to a 'census', if you will.

As you can see, the total population for December was 50, and the population this far in January is 60.

Count	Date
15	12/1/2018
16	12/15/2018
19	12/30/2018
35	1/1/2019
25	1/9/2019

Losses is just a table detailing events related to that population.

As you can see, there were 10 in December and 6 so far in January.

ID	Date
Ashely	12/4/2018
Naida	12/6/2018
Violet	12/13/2018
Pa	12/14/2018
Alline	12/16/2018
Jesusa	12/16/2018
Lucrecia	12/18/2018
Bronwyn	12/20/2018
Gene	12/21/2018
Velda	12/26/2018
Ken	1/1/2019
Rossie	1/1/2019
Olevia	1/3/2019
Patricia	1/5/2019
Carylon	1/9/2019
Bob	1/10/2019

BI-DATE_DIM is simple a calendar table. It has 1:* relationships with the other tables with dates. You don't need to know why this is, but it is necesary.

DIM_DATE

12/1/2018

12/2/2018

12/3/2018

12/4/2018

12/5/2018

12/6/2018

12/7/2018

12/8/2018

12/9/2018

12/10/2018

12/11/2018

12/12/2018

12/13/2018

12/14/2018

12/15/2018

12/16/2018

12/17/2018

12/18/2018

12/19/2018

12/20/2018

12/21/2018

12/22/2018

12/23/2018

12/24/2018

12/25/2018

12/26/2018

12/27/2018

12/28/2018

12/29/2018

12/30/2018

12/31/2018

1/1/2019

1/2/2019

1/3/2019

1/4/2019

1/5/2019

1/6/2019

1/7/2019

1/8/2019

1/9/2019

1/10/2019

Given these numbers, we would expect to get:

LM = 10 / 50 * 365 / 31 = 235%

Why not 20%? because I am required to use annualized numbers.

[jengwt]

@v-yuezhe-msft I would like to state that just because I found a workaround doesn't mean the root problem is resolved. I disagree with the status being "delivered", and suggest that Microsoft still take this up as a bug.

[jengwt]

@v-yuezhe-msft I would have, but it wouldn't let me mark my own reply as a solution. Thank you for doing so.

[jengwt · Accepted solution]

Ok I figured it out. It's actually pretty simple; you just explicitely tell PBI what to do in January:

Population_LM =
VAR LMDs = [LM_Days]
VAR DN = [DNOW]
VAR MDN = MONTH(DN)
VAR YDN = YEAR(DN)
RETURN ROUND(
    DIVIDE(
        CALCULATE(
            SUM('Population'[COUNT])
            , CALCULATETABLE('Population'
                , MONTH('BI-DATE_DIM'[DIM_DATE]) = 
                        IF(MDN = 1
                            , 12
                            , MDN
                        )
                    && YEAR('BI-DATE_DIM'[DIM_DATE]) = 
                        IF(MONTH(DN) = 1
                            , YDN - 1
                            , YDN
                        )
            )
        )
    , LMDs
    )
, 0
)

and,

Loss_LM = -- STATIC CALC
VAR DN = [DNOW]
VAR MDN = MONTH(DN)
VAR YDN = YEAR(DN)
RETURN CALCULATE(
    COUNT('Losses'[ID])
    , CALCULATETABLE(Losses'
        , MONTH('BI-DATE_DIM'[DIM_DATE]) = 
                IF(MDN = 1
                    , 12
                    , MDN
                )
            && YEAR('BI-DATE_DIM'[DIM_DATE]) = 
                IF(MDN = 1
                    , YDN - 1
                    , YDN
                )
    )
)

I still think those PREVIOUS...() functions don't really work as envisioned.

[jengwt]

@v-yuezhe-msft It will be difficult to resolve this after January and the functions resume working "correctly" again.

Want to try having a call?

[jengwt]

@v-yuezhe-msft Literally all I did was swap in those functions. For example: 

Population_LM =
VAR LMDs = [LM_Days] 
--VAR DN = [DNOW] 
RETURN ROUND(CALCULATE(SUM('Population[COUNT])
, PREVIOUSMONTH('BI-DATE_DIM'[DIM_DATE]) 
/* CALCULATETABLE('BI-DATE_DIM' 
, MONTH('BI-DATE_DIM'[DIM_DATE]) = (MONTH(DN) - 1)
&& YEAR('BI-DATE_DIM'[DIM_DATE]) = YEAR(DN)
) */
) / LMDs, 0)

'BI-DATE_DIM'[DIM_DATE] is a mm/dd/YYYY field.

Before:

After:

[v-yuezhe-msft]

@jengwt,

Yes, these functions will return blank value if you don't use date type fields(Date, Year, YearMonth,etc) to filter them. You can include yearmonth slicer in your report.

Regards,
Lydia

[jengwt]

@v-yuezhe-msft I don't understand, are those functions somehow dependent on YearMonth? And when I modify my report to include PreviousMonth() and PreviousYear(), the measures return (BLANK).

[v-yuezhe-msft]

@jengwt,

The PREVIOUSMONTH(), PREVIOUSQUARTER(), and PREVIOUSYEAR() functions will return correct result with date context. If you don't want to put yearmonth column in the Matrix, you can create a slicer using the yearmonth column and select value in the slicer.

Regards,
Lydia

[jengwt]

@v-yuezhe-msft I tried modifying my report to use these calculations, and not only do they still yield (Blank), they actually they break my existing calculations which worked using my code.

Seems like PREVIOUSMONTH(), PREVIOUSQUARTER(), and PREVIOUSYEAR() don't work either.

What is the point of YearMonth? What do you do with that? Oh, that's because it's the determining row in the matrix? Unfortunately, we're not displaying the data in that way.

[v-yuezhe-msft]

@jengwt,

I create the following measures in your date table.

LASTCOUNT = CALCULATE(COUNT(Losses[ID]),PREVIOUSMONTH(DIM_DATE[DIM_DATE]))

LASTSUM = CALCULATE(SUM(Population[Count]),PREVIOUSMONTH(DIM_DATE[DIM_DATE]))

LM = [LASTCOUNT]/[LASTSUM]*365/31

And create the following column in the date table.

YearMonth = FORMAT(DIM_DATE[DIM_DATE],"YYYY-MM")

Regards,
Lydia

[jengwt]

@v-yuezhe-msft I posted some sample data to the OP.

[v-yuezhe-msft]

@jengwt,

Could you please share sample data of your original table and post expected result here?

You can follow the guide in the blog below to share data.
https://community.powerbi.com/t5/Community-Blog/How-to-Get-Your-Question-Answered-Quickly/ba-p/38490

Regards,
Lydia