Structural sampling in the field is usually dependent of the mapping path chosen by the mapper. Different mapping schools map and sample structural data either across structures or along form surfaces in a study area. The collected structural data are either dense or sparse which often depends on the availability and quality of the outcrop. These outcrop-scale observations may not be perfectly representative of the larger 3D geometry. For example, measurements may not capture the full wavelength of a sampled fold. In this study, we assess the effect of structural data distributions on modelled 3D fold geometries using the new open-source modelling engine LoopStructural. LoopStructural models folds within a fold frame which coordinates correspond to the axes of the finite strain ellipsoid. The geometry of the modelled fold profiles is analysed using the S-Plot which plots the fold limb and fold axis rotation angles against the z and the y fold frame coordinates. The rotation angles are used in the modelling framework of LoopStructural to characterise the folded surfaces. The wavelength of the fold is analysed using the S-Variogram (A sample semi-variogram). The workflow of the experiment consists of 1) generating single fold and interference pattern 3D models, 2) sampling structural data from the reference models: i) randomly, ii) across structures (across the direction of the axial surface) and iii) along the trace of the folded foliation (similar to form surface mapping); 3) build 3D models from these new samples, and 4) compare the resulting models against each reference fold geometry to assess the recovery of the fold wavelength and the effects of the across and along structure sampling on 3D fold geometries. The model recovery of wavelength is assessed using a Spacing Ratio of data spacing of each dataset to the reference wavelength. Data spacing is the distance between a given data point and its nearest neighbour. We present results on the assessment of the minimum amount of structural data needed to recover the fold wavelength testing random samples. The modelled 200 random datasets of cylindrical tight symmetric folds and asymmetric folds show that the datasets that best recovers the wavelength of the reference model have a Spacing Ratio of roughly 0.02 with a mean of data spacing of 400m for a reference wavelength of 20km.