Maybe this will help someone. You can return the favor by telling me how I can do this in R!
I wrote a Calculated Column that returns the change in a value compared to the prior period, while ensuring a few conditions are met. The table contains 75,330 records of property management data for approximately 100 different nonprofit organizations submitted over 71 reporting periods. I didn't use any Time Intelligence functions because there are eleven different fiscal year-end's and I didn't know where to begin. The table has 18 total columns; but, here's the general structure:
Organization PropertyID Months PeriodNumber PeriodID NCF ... Org1 abc 3 20031 18 2200 ... Org1 abc 6 20032 19 4600 ... Org1 abc 9 20033 20 6700 ... Org1 abc 12 20034 21 8000 ... Org1 abc 3 20041 22 4000 ... Org1 abc 6 20042 23 12000 ...
As you can see, the data are reported cumulatively (3 months this quarter, 6 months next quarter, etc.), and I needed to calculate the change from the prior-period so all of the records represent just three-months of property management history. Unfortunately, the reporting window is based on the calendar year, not each organization's fiscal year-end (i.e. 3-months reported on 3/31, 6-months on 6/30, etc.). The PeriodID column ranges from 1 to 71, and it's simply an index column based on the unique values of PeriodNumber sorted in ascending order. PeriodNumber is of the format, YYYYQtr (so Q1 2004 is 20041). As I mentioned, these 100 organizations share 11 different fiscal year-end's, and my data goes back to 2001. Here's the Calculated Column:
ChgNCFfromPP =
IF (
ISBLANK ( QuarterlyIndicators[NCF] ),
BLANK (),
IF (
QuarterlyIndicators[Months] = 3,
QuarterlyIndicators[NCF],
IF (
ISBLANK (
LOOKUPVALUE (
QuarterlyIndicators[NCF],
QuarterlyIndicators[PropertyID], QuarterlyIndicators[PropertyID],
QuarterlyIndicators[PeriodID], QuarterlyIndicators[PeriodID] - 1
)
),
BLANK (),
QuarterlyIndicators[NCF]
- LOOKUPVALUE (
QuarterlyIndicators[NCF],
QuarterlyIndicators[PropertyID], QuarterlyIndicators[PropertyID],
QuarterlyIndicators[PeriodID], QuarterlyIndicators[PeriodID] - 1
)
)
)
)
First, if the reported NCF in the current period or the priod period is blank, a blank is returned. Second, if the number of months reported = 3 in the current period, the current period's NCF is returned (because the goal was to get 3-months of data for each period).
Here's the output:
Organization PropertyID Months PeriodNumber PeriodID NCF ChgNCFfromPP FiscalYear Org1 abc 3 20031 18 2200 2200 2003 Org1 abc 6 20032 19 4600 2400 2003 Org1 abc 9 20033 20 6700 2100 2004 Org1 abc 12 20034 21 8000 1300 2004 Org1 abc 3 20041 22 4000 4000 2004 Org1 abc 6 20042 23 12000 8000 2004
I can now group by Organization and FiscalYear, and the values for NCF - and any other variables in which I apply the same formula - will represent 12-months of reporting history.
I searched all over the place for a way to deal with multiple fiscal year-end's, and I didn't have any luck. So, I'm sharing this to hopefully save someone else some time in the future. Or, suggest a way it can be improved!