Possibly it is too obvious, but you could point out that
merge commands are typical steps in which there is a chance that you (significantly) reduce your sample because the information you wish to merge is not available for every observation unit. merge should to be applied with care and lost observations should be tracked.
merge can also be applied to explore (1) how many observation units are in dataset A and in B or (2) how many observation units are in dataset A but not in B. This allows to, for example, learn about selection in the dataset.
Possibly it is too obvious, but you could point out that
merge commands are typical steps in which there is a chance that you (significantly) reduce your sample because the information you wish to merge is not available for every observation unit. merge should to be applied with care and lost observations should be tracked.
merge can also be applied to explore (1) how many observation units are in dataset A and in B or (2) how many observation units are in dataset A but not in B. This allows to, for example, learn about selection in the dataset.