IR radiation goes further into a material than does light in the visual spectrum, so it tests a little deeper than just the surface. How deep into a material the radiation penetrates depends on the material however, so I think it may be hard to give a simple answer. A black bag probably won’t work (stuff that is black in the visual spectrum tend to also absorb IR radiation well) and an orange peel is probably too thick if you are hoping to say anything about e.g. the pulp (however, the peel of an orange may have markers that correlate well to qualities of the pulp, so this is not as futile as trying to determine properties of something inside a black bag).
A little primer on NIRS: https://en.wikipedia.org/wiki/Near-infrared_spectroscopy
Otto is right. Penetration depth heavily depends on the sample – both material and density.
Usually, SCiO penetrates food for 3-4 mm and less than 1 mm for pills (which have compressed powder in them).
While in principle SCiO can penetrate wrapping such as the plastic bags, individual scans will depend on material type, thickness, and transparency.
We recommend scanning the object directly to get the most accurate read.
Regarding fruit and vegetable peels, we are able to scan accurately through vegetables and fruits with a thin rind or skin (such as apples, pears, cucumbers etc.) since the illumination beam light is still returned to the sensor and our algorithms eliminate their effect. In some cases we are also able to scan and return data regarding fruits with a thick rind, such as melon and watermelons, because to an extent, the content of the rind correlates to the content of the actual fruit.