See Jerkstrand et al 2015 (Supersolar Ni/Fe production in the Type IIP SN 2012ec).
The nice thing about this method is that its quite robust to several uncertainties. Because of the similar ionization potentials of Ni I and Fe I, its likely that the Ni II/Fe II ratio reflects the Ni/Fe ratio (and this is supported by the full spectral models which for most regimes predict that most Ni and Fe is in the singly ionized form). Because of the similar excitation potentials, the temperature sensitivity is weak and thus even significant errors in T does not propagate to a big error in the abundance ratio. And, when NIR spectra are available, the conclusion can be independently tested using [Ni II] 1.9 mu.
This ratio also turns out to be a really constraining diagnostic for which layer in the progenitor star was explosively burnt (see our follow-up paper).
The method assumes that the lines are optically thin: in a standard Type II model at 370d the optical depths are 0.04 in [Ni II] 7378 and 0.08 in [Fe II] 7155. Roughly, these optical depths scale with M*V^(-3)*t^(-2). E.g. a narrow-lined SN may have 8 times lower Ni and Fe masses, 2 times lower V, so the optical depth is similar for the same epoch. But beware for e.g. ECSNe and ECSN-like explosions the Ni mass may be higher than in a normal iron core collapse, while velocities lower - in the early nebular phase the lines are then optically thick.
The other thing to watch out for it contamination by primordial Ni and Fe emission. The lower the amount of explosive nucleosynthesis, the more severe this contamination may be because the amount of primordial metals does not change much with MZAMS. Therefore, it is harder to do this for SNe with very low 56Ni masses. Also, the later you wait the more will gamma rays in general escape from the explosive burnt region (so powering of the primordial stuff relative to the synthesised increases). In addition, at too late phases NLTE effects becomes stronger. While one of the strengths of the method is that departure coefficients are expected to be similar, this nevertheless increases uncertainty.
In summary: the method works best at an intermediate nebular phase - too early and lines may be optically thick still, too late and there will be increased contamination by primordials in addition to NLTE effects. For a normal IIP model the sweet-spot is 300-400d.