Understanding the $\mu$ problem I am going through the motivation to introduce singlet extensions of the MSSM (like for instance the NMSSM), and of course such a motivation is naturalness. Here, one argues that the $\mu$ parameter could not be zero ( for phenomenological reasons, extra unacceptable symmetries ...), nor too large (in order to give rise to SSM and hence masses...), but it must be of the order of $M_{SUSY}$. My question is quite simple:
Why a natural value for $\mu$ would be at the Planck (or GUT) scale?
 A: In the Standard Model, the natural scale for the dimension-two Higgs coupling, $m_H^2$, is the Planck scale, because it takes radiative corrections $\delta M_H^2 \sim M_P^2$. 
The $\mu$-parameter in the MSSM (or any softly-broken SUSY theory), however, is protected from quadratic divergences by supersymmetry (see the non-renormalization theorem). There is no natural scale for the $\mu$-parameter (although it might be argued that if it originates from unknown high-energy physics, it ought to be close to the scale of that high-energy physics).
The $\mu$-problem isn't as severe as the hierarchy problem in the SM. If $\mu$ could be anywhere between, say $M_W$ and $M_P$, stable under radiative corrections, and is unrelated to SUSY breaking, why should it be that $\mu\sim M_{SUSY}$? That's the $\mu$-problem. In the MSSM, it is a necessary accident. It is necessary because it is required by the EWSB scale, $M_Z^2 \sim -\mu^2 + m_{SUSY}^2$.
In NMSSM models that solve the problem, there aren't two scales ($\mu$ and $M_{SUSY}$), there is simply one scale, $M_{SUSY}$, which results in a singlet VEV, $\langle s\rangle\sim M_{SUSY}$. An effective $\mu$-term comes from an $SH_u H_d$ operator, with $\mu_{eff} \sim M_{SUSY}$. A tree-level $\mu$-term, present in the MSSM, is forbidden by a discrete symmetry (often a  $Z_3$).
Thus, there is a mechanism and a symmetry which insures $\mu\sim M_{SUSY}$ -it is not an improbable coincidence as it was in the MSSM.
