Hello @egobbo .
You can also create calculated fields at the dataset level and then create a filter at the dataset level as well.
This way you can filter the data directly there and make sure that all the analyses that use it received the filtered data.
However remember that it is the best practice to work with already clean and massaged datasets to avoid overloading the BI layer with calculations and filters so your visuals can render blazing fast.
I would recommend to create a data pipeline using AWS Glue or or your favorite ETL tool to ensure that the data that arrives to the consumption layer (Quicksight in this case) doesn’t contain duplicates.
Hope it helps, please mark this solution as solved if that’s the case also to help other members of the community., otherwise let us know.
Happy dashboarding!