Observation allows first hand exposure to the experiment or behaviour under examination, rather than using self-reporting methods or surveys which may become skewed.
Recording actual behaviour is a lot more scientific.
Down side to the observation method
The negatives are pretty straight forward: Running this type of analysis takes time. There's also the fact that, people may behave in a different way if they know they are being observed, and often consent to observe is a prep-requisite that invalidates the experiment or learning.