I made some progress with my previous question, but here is a follow up: I am trying to build a table that derives the balance of an account at the beginning and end of a month (month is an example, this should also work on any other time granularity). Given is a table with transactions and each transaction has a balance before and after the transaction, Transactions have a transaction number that is increasing but can have gaps (similar to a rank). I already managed to build metrics that compute the last transaction before a month and the first transaction after a month. Now I want to build a metric that simply uses the balance before / after that transaction, but I don’t manage to do it, because I have to look across the granularity of the data.

Hi Thomas, There is a bit of a limitation in the GD Platform when trying to force null values in dates in running tables. Unfortunately, the only work around is to have those exact dates placed in your source file. Your raw data needs to contain all the dates that include the NULL values as well. This is the only way to make the calculations work on a running date total. If you are trying to connect based on other datasets, you can define this in a metric using the “BY” clause. This can still be restricted by how your LDM has been constructed, but the basic idea of this can be found in a legacy article called, Explicit Lifting. Please note that the images are out of date, but the theory is the same in connecting datasets.

Hi Joseph, thanks for that insight. Now how would we do it on a dataset that connects to lots of different dimensions? Each filter variant would remove rows and potentially also dates, right? If we include a NULL entry for every Attribute / Value combination out there, we’d introduce billions of useless entries. Its basically not possible

Hi Thomas, I’m afraid that it isn’t that simply within the platform. We can raise this as product feedback, but I’m afraid that it wouldn’t necessarily be considered as working in the LDM like this could lead to terrible performance issues.