The food, beverage and natural supplements industries face many challenges relating to quality control as influenced by natural product variation, formulation errors and product adulteration. Conventional chromatographic and spectrophotometric assays can be time-consuming, costly and insensitive to key quality parameters and adulterants. Thus, there is a recognized need for a rapid, sensitive, non-destructive optical technique. This study investigates optimization of multivariate calibrations of chemical product compositions analyzed using A-TEEM spectroscopy. The A-TEEM method provides rapid (s-min) analyses of all chromophoric and fluorescent compounds in the UV-VIS range with micro- to sub-microgram/L sensitivity to many aromatic compounds. Conventional fluorescence EEM analyses primarily focus only on the inner-filter-effect corrected fluorescence data. In this paper, we systematically investigated the statistical significance of evaluating the separate absorbance, EEM and combined multi-block absorbance and EEM data variable sets. We report that there is a consistent improvement for the majority of A-TEEM models using ‘multi-block’ data organization compared to the separate absorbance or EEM variable data sets. We attribute the increased statistical advantage to the fact that most chemicals, even with similar structures, in a given solvent exhibit unique molar extinction coefficients, fluorescence quantum efficiencies as well as absorbance- and fluorescence excitation and emission spectral shapes. The A-TEEM data analyses further compared the Extreme Gradient Boosting (XGB) algorithm for both discrimination and regression to other comparable methods including Partial-Least Squares and Support Vector Machine. The conclusion, based on a multi-instrument comparison, was that the XGB analyses of the multi-block A-TEEM data lead to the most effective validation for both discrimination and regression models of food, beverage and natural supplements chemical composition and adulteration.