How would you handle a dataset where 20% of the lead source data is missing or corrupted for Tessian?
