A key first step in natural product structure elucidation is a determination of whether a newly isolated compound is a known or unknown one, a process called dereplication. Both ¹H and ¹³C databases can be used for this purpose and we discuss the merits and demerits of each. Our conclusion is that the most effective database for dereplication should be based on edited HSQC data in tabular form. A recent program called SMART demonstrates the potential of this approach and much of the needed data for many additional compounds are already available. The more general question of how to archive all NMR data is also discussed, based on several recent reviews of the subject. The advantages and limitations of archiving raw FID data are discussed, along with suggestions of how such a database could be maintained.