On one hand I was planning to work on all data formatting and calculated fields on the dataset level, so it can be consumed by different analytics without the need to implement them al over again in each analytics.
However, I have notices that when doing so, some records can be skipped on dataset refresh, and it will impact the measures and analytical results.
I would expect the formatting to skip the problematic values instead of filtering out entire records from the dataset.
@Annab - Formatting has nothing to do with record skipping. I believe your records are getting skipped due to some metadata related issues, whereas formatting is entirely related to representation. In case, you want all your records to be loaded perfectly each time into SPICE dataset then you have to ensure the correct metadata of the source data. Otherwise, you can try to go for direct query and see if the rows that were getting dropped during SPICE load can be seen in the report/dataset. Hope this helps!