The implicit unsteady approach is appropriate if the time scales of the phenomena of interest are of the same order as the convection and/or diffusion processes (for example, vortex shedding) or are due to some relatively low frequency external excitation (for example, time-varying boundary conditions or boundary motion)
In the implicit unsteady approach each physical time-step involves some number of inner iterations to converge the solution for that given instant of time. These inner iterations may be accomplished using the same implicit integration or explicit integration schemes used for steady analysis. The physical time-step size used in the outer loop is specified by you, whereas the inner iterations are marched by the integration scheme using optimal local steps as determined by the Courant number.
With implicit unsteady approach you are required to set the physical time-step size, the Courant number, and the number of inner iterations to be performed at each physical time-step.
The physical time-step size will generally be governed by the transient phenomena being modeled. The time-step should at least satisfy the Nyquist sampling criterion; more than two time-steps per period are required.
In general the same Courant number setting guidelines as for the steady-state integration schemes apply.
The number of inner iterations per physical time-step is harder to quantify. Generally, this will need be determined experimentally. You should select a certain number of inner iterations (the default is 20) and then see if the results are affected by increasing or decreasing this number.
Smaller physical time-steps generally mean the solution is changing less from one time-step to the next, so that fewer inner iterations are required. There will be an optimal balance of time-step size and number of inner iterations for a given problem and desired transient accuracy.