The linear [ χ (1) ] and second-order nonlinear [ χ (2) ] optical susceptibilities of the 2-methyl-4-nitroaniline (MNA) crystal are calculated within the local fi eld theory, which consists of fi rst computing the molecular properties, accounting for the dressing e ff ects of the surroundings, and then taking into account the local fi eld e ff ects. Several aspects of these calculations are tackled with the aim of monitoring the convergence of the χ (1) and χ (2) predictions with respect to experiment by accounting for the e ff ects of (i) the dressing fi eld within successive approximations, of (ii) the fi rst-order ZPVA corrections, and of (iii) the geometry. With respect to the reference CCSD-based results, besides double hybrid functionals, the most reliable exchange-correlation functionals are LC-BLYP for the static χ (1) and CAM-B3LYP (and M05-2X, to a lesser extent) for the dynamic χ (1) but they strongly underestimate χ (2) . Double hybrids perform better for χ (2) but not necessarily for χ (1) , and, moreover, their performances are much similar to MP2, which is known to slightly overestimate β , with respect to high-level coupled-clusters calculations and, therefore, χ (2) . Other XC functionals with less HF exchange perform poorly with overestimations/underestimations of χ (1) / χ (2) , whereas the HF method leads to underestimations of both. The fi rst-order ZPVA corrections, estimated at the B3LYP level, are usually small but not negligible. Indeed, after ZPVA corrections, the molecular polarizabilities and fi rst hyperpolarizabilities increase by 2% and 5%, respectively, whereas their impact is magni fi ed on the macroscopic responses with enhancements of χ (1) by up to 5% and of χ (2) by as much as 10% − 12% at λ = 1064 nm. The geometry plays also a key role in view of predicting accurate susceptibilities, particularly for push − pull π -conjugated compounds such as MNA. So, the geometry optimized using periodic boundary conditions is characterized by an overestimated bond length alternation, which gives larger molecular properties and even larger macroscopic responses, because of the local fi eld factor ampli fi cation e ff ects. Our best estimates based on experimental geometries, charge dressing fi eld, ZPVA correction, and CCSD molecular properties lead to an overestimation of χ (1) by 12% in the static limit and 7% at λ = 1064 nm. For χ (2) , the di ff erence, with respect to the experiment, is satisfactory and of the order of one standard deviation.