Forum Discussion
Calculate Difference from Previous 'Date' Dynamically
Cool question - using the query editory you can sort the table and add an index so that each row has a distinct number (0,1,2,3, etc). Here's the formula you want to write ... I broke it out into pieces.
Total = SUM(Total) Total - Previous Period = VAR thisperiod = MAX(Index) RETURN CALCULATE ([Total], FILTER (ALL(Table), Index = (thisperiod - 1)) Total - PoP Change = IF( [Total] && [Total - Previous Period], [Total] - [Total - Previous Period] ) Total - PoP Change - Aggregates Correctly = SUMX(Table, [Total - PoP Change] )
With current and previous date calculations we usually use a date table and the date/time functions but those calculations don't work with weeks, so you'd have to do it this way.
No that makes sense - should work fine, here's what you do.
Option 1
- Undo the sorting and indexing.
- Right click your query and hit duplicate.
- In this new query you're going to remove all the columns except the week, then remove the duplicates, sort the column, and add the index.
Option 2 - This is a continuation of Option 1, You can keep two tables (option 1) or put the index back on the original query (option 2), your choice.- Right click the new query and uncheck the "Enable Load" option
- In the old query, select the weeks column and click on the merge queries option up in the ribbon. Merge the old query with the new query based using the week column. Left Join is chosen by default - that's what you want.
- Expand the new column to show the index column.
All measures - the last one gives us the right number for the grand total (i hope that's what it does!)
Cool question - using the query editory you can sort the table and add an index so that each row has a distinct number (0,1,2,3, etc). Here's the formula you want to write ... I broke it out into pieces.
Total = SUM(Total) Total - Previous Period = VAR thisperiod = MAX(Index) RETURN CALCULATE ([Total], FILTER (ALL(Table), Index = (thisperiod - 1)) Total - PoP Change = IF( [Total] && [Total - Previous Period], [Total] - [Total - Previous Period] ) Total - PoP Change - Aggregates Correctly = SUMX(Table, [Total - PoP Change] )
With current and previous date calculations we usually use a date table and the date/time functions but those calculations don't work with weeks, so you'd have to do it this way.
Thanks for your solution,
However I did think I included this in original description but looks like I missed it out,
As it was a simplified table, I missed out the fact that there are several different customers also included in the table in a separate column. Therefore there for multiple occurrences of the same week for instance, therefore my data would actually displayed more like the below screenshot after applying the sort and indexing: (Before and after sorting by Week and adding index)
Also, ontop of that, there is another column for sub categories of customer, so there could be for instance 2 rows for WK01 Customer A (each sub category).
Would your solution also work with this format of data as it would be looking at index 0 and 1 for instance, which is 2 individual customers but same week.
Sorry if I'm not making much sense!
Thank again
No that makes sense - should work fine, here's what you do.
Option 1
- Undo the sorting and indexing.
- Right click your query and hit duplicate.
- In this new query you're going to remove all the columns except the week, then remove the duplicates, sort the column, and add the index.
Option 2 - This is a continuation of Option 1, You can keep two tables (option 1) or put the index back on the original query (option 2), your choice.- Right click the new query and uncheck the "Enable Load" option
- In the old query, select the weeks column and click on the merge queries option up in the ribbon. Merge the old query with the new query based using the week column. Left Join is chosen by default - that's what you want.
- Expand the new column to show the index column.
I have now tested the same solution following your steps for multiple customers (extra column that will be filtered on)
Finding that there is an issue with the formula reading previous date relevant to that filter
What appears to be happening... When there is no filter selected the values show correct, i.e. overall total shows overall difference by Week. No problems.
However, when applying a filter, it appears as though it is picking up the relevant quantity for the customer/week, but when it is trying to take away from the previous week, it is calculating Customer A week 2 / Customer B week 2 / Customer C week 2, then taking Customer C week 1 away for instance.
Example: As you can see from the screenshots above,
Customer A week 1 = 5
Customer B week 1 = 4
Customer C week 1 = 6
Total 15
Therefore, the graph is showing Customer C week 2 (3) - 15 to result in -12
Thanks for your assistance on this :smileyhappy:
austinsense these are all measures right?
What does the last one do?
Total - PoP Change - Aggregates Correctly = SUMX(Table, [Total - PoP Change] )
All measures - the last one gives us the right number for the grand total (i hope that's what it does!)
Eventually got my head round out it!
Solution works perfectly, thanks for your detailed descriptions
Thanks!
