Practical - Compared to participant observation methods, it reduces the risk of HAWTHORNE EFFECT. If people do not know they are being watched, they will not change the way they behave.
DISADVANTAGES
Theoretical - Depending on the method, NPO can be highly structured and used to produce quantitative data or qualitative data - whatever data produced the researcher will not be able to gain meaning/understanding of why events are happening, reduce validity of data.
Ethical - If people don't know that they are being observed they are being deceived and the researcher has not gained informed consent - however, by doing so they run the risk of altering behaviour.