Mid-infrared spectroscopy (MIRS) was used to simultaneously detect and predict concentrations of D-sorbitol and L-sorbose during a Gluconobacter suboxydans biotransformation. Quantitative models for both these compounds were developed for the entire time-course of the process and validated externally using samples not included in the original modelling exercise, giving standard errors of prediction of 3.29 and 3.3% for sorbitol and sorbose, respectively, and a correlation coefficient close to 1.