Fivetran Mongo connector is becoming very expensive since it does not implement the partitioned table functionality from Bigquery to save the query cost. The partitioned field in Bigquery is often a timestamp, date, or an integer range. Currently, Fivetran only reads the primary key which as a result it has to always scan the whole table to update the data. The connector should be able to filter the right data point as well as the partition it's in.