Comment by RandomLensman

Not always/limited in some areas (astrophysics come to mind where we, e.g., cannot (yet) create a star under controlled conditions). Testing hypotheses or predictions vs observations is also a valid method.