In the present work, a near-infrared spectroscopy (NIRS) methodology has been developed and tested for binary flows of granular lactose/microcrystalline cellulose (MCC) mixtures. Powder flow is characterized in a guillotine-like sampling system, making radial and axial sampling of flow possible at regular intervals. The guillotine's operation parameters are first optimized through statistical analysis to minimize errors due to phenomena occurring between the guillotine's surface and the powder (e.g. drag, backflow). The experimental design confirms the influence of guillotine speed and particle size of the powder on powder flow. While it is clear that this invasive system will cause radial disturbances, they have been minimized (relative error less than 2.83%) and their presence can be quantified appropriately for sufficiently accurate sampling. The proposed guillotine-based NIRS methodology is able to efficiently and reproducibly measure the average residence time (tau) of the pulsed material (MCC) and its relative standard deviation (sigma) as a function of radial position. The results also demonstrate that sigma is positively correlated to particle size ratio (MCC/lactose ratio between 1:1 and 1:3) and that, consequently, both the speed of bulk dispersion and the speed of natural convection of the dispersed phase are faster with higher particle size ratios, leading to the conclusion that the system is less stable and that segregation tendency is higher. (C) 2012 Elsevier B.V. All rights reserved.