The purpose of this study was to use a near-infrared (NIR) spectroscopic sensing system to monitor each cow's milk quality accuracy every 20 seconds and every time the cow was milked. We used the milk fat, lactose, milk urea nitrogen (MUN), and somatic cell count (SCC) as four indices of milk quality. Raw milk samples were obtained from four Holstein cows belonging to Hokkaido University. Using an NIR sensing system, raw milk's NIR spectra were recorded every 20 seconds while the cows were being milked. The wavelengths of the spectra ranged from 700 to 1050 nm. Using the MilkoScan instrument, milk fat, lactose, and MUN were measured while a Fossomatic instrument was used to measure SCC. Partial least squares regression analysis was used to generate calibration models in order to verify the precision and accuracy of the models. The obtained results demonstrated that the accuracy of each cow's milk quality measurement every 20 seconds and at one milking time during the milking process was outstanding for milk fat and comparable for milk lactose. For every cow, the MUN and SCC findings showed a considerable difference. These findings showed that the measurement of each cow's milk quality could have an effect on the calibration models' precision and accuracy.