Simulated moving-bed chromatography (SMBC) separation of a solution containing three different saccharides was investigated by real-time, inline monitoring of the concentration of each saccharide with Fourier transform near-infrared spectroscopy (FT-NIRS). We built partial least squares (PLS) regression models to estimate the saccharide concentrations based on preprocessed spectral data in the wavenumber range 6,100?5,440?cm?1 that excludes the water absorption band. To verify the calibration model, we performed separation of the raw material solution of the test set using SMBC, and confirmed that the predicted concentration of each saccharide from the calibration model corresponds well with the actual measured concentration of each saccharide obtained by HPLC. The collection range can be adjusted online using the monitoring values in order to maximize the purity and yield of the target saccharides.