Hi Gabriel,
yes the RowNum would be the correct approach.
The item you want to group by is the column mentioned in PartitionBy.
i.e. for you it would count the row number per IdClient.
Note: the result will contain all rows still.
To keep only the highest number, you need to add a deduplication task, choose Expression as method, choose the calculated field containig the row number and choose to keep the highest value