IBM Interview Question

How would you handle missing or inconsistent data in a dataset you're analyzing?