Near infrared (NIR) spectroscopy was used to determine the drug concentration in 3% (w/w) acetaminophen blends within the complex flow regime of the tablet press feed frame just before tablet compaction. NIR spectra also provided valuable information on the powder flow behavior within the feed frame and were used to track when a process enters or leaves the steady state. A partial least squares regression calibration model was developed with powder mixtures that varied from 1.5 to 4.5% (w/w) by obtaining 135 spectra after steady state for each concentration while the feed frame and die disc operated at 30.5 revolutions per minute (rpm). The calibration model determined drug concentration in validation blends with a root mean square error of prediction and bias below 0.1% (w/w). The robustness of the NIR calibration model was evaluated by determining the effect of variation on the operating conditions (paddle wheel speed and die disc speed) on NIR predictions. This work found that the paddle wheel speed can be increased up to 30% and the die disc speed decrease 10% without affecting NIR predictions. The results demonstrated that paddle wheel speed has a significant effect on the wave powder behavior (frequency and amplitude) but does not have significant effect on the mass hold-up within feed frame. The die disc speed does not affect wave powder behavior but affects significantly the mass hold-up inside the feed frame. This information can be used to reduce the tablet weight variability and ensure that this critical attribute is met.