I think, because if you know frequency response of the system, you can easily predict its response in time domain for any type of input signal. (And vice-versa as well). Another reason could be well known technique of determining system stability based on frequency response.
This must be not a simplest explanation, but anyway...
Every system has a transfer function in frequency domain: T(ω). Very often it is done not as a function of frequency ω (ω=2πf), but Laplace parameter s=jω: T(s).
If you apply signal f(t) to the system and want to know output g(t), there is a simple formula in frequency domain:
G(s) = T(s)*F(s),
where G(s) and F(s) are Laplace transforms of g(t) and f(t).
So, you have f(t), then found F(s), multiply by T(s), get G(s), performe inverse Laplace transform to obtain g(t).
nl5 said:
Another reason could be well known technique of determining system stability based on frequency response.
This is part of control theory. Simple example is using gain and phase margins on Bode plot to analyse stability of the closed loop system. Here is a link.