A value is far from the rest of the data. What should a data scientist do first?

A. Delete it automatically B. Treat it as proof of fraud C. Replace the entire column with its mean D. Investigate whether it is an error, a rare valid case, or evidence of a distinct process