A successful transition from a laboratory proof-of-principle to a functioning industrial Process Analytical Technology (PAT) application of spectroscopy and chemometrics is still an active area of research and development. A comprehensive understanding on how the design and implementation of the optical instrumentation affect the data quality, and how this will affect the performance of the prediction models is vital for the successful implementation of in-line monitoring. In a previous study, we have demonstrated that near-infrared spectroscopy (NIRS), combined with chemometric techniques, can be used to quantify α-lactalbumin and β-lactoglobulin in aqueous whey solutions. This work demonstrates the potential, both at-line and in-line, of monitoring a protein fractionation process in a full-scale production. An in-line near infrared spectrometer (NIRS) was used to monitor the individual protein concentrations in a protein fractionation process over 20 days. In addition, samples were extracted from the process and analysed with at-line NIRS and an in-house HPLC reference method. The developed models could predict β-lactoglobulin and α-lactalbumin concentrations with satisfactory precision and accuracy, yielding a root mean square error of cross-validation of 0.08 and 0.18 w/w% proteins, respectively. For the first time, the whey proteins concentrations were measured continuously in a production facility to demonstrate the potential of NIRS for at-line and in-line rapid monitoring of protein composition in industrial whey streams. A PAT tool was developed from NIRS data where Partial Least Squares regression modelling and an Exponentially Weighted Moving Average filtering were used to extract and visualize valuable information on the process dynamics and performance. The study demonstrates that continuous monitoring by in-line NIRS allowed elucidation of some process behaviours not readily detectable with the current sampling frequency and that it offers novel process optimisation opportunities by providing increased process understanding and decision support information concerning preventive maintenance and real-time process validation.