Front-face fluorescence spectroscopy was used to characterize 16 ice cream formulations (four fats, two mono-/diglyceride mixtures, and two protein mixtures). Tryptophan fluorescence spectra were recorded directly from the ice cream samples. The discriminant ability of the data was investigated by principal component analysis (PCA) and partial least square (PLS) discriminant analysis. The similarity maps defined by principal components 1 and 2 allowed clear discrimination between the protein mixtures. The spectral pattern associated with protein PLS discrimination provided characteristic wavelengths that were the most suitable for separating the spectra. Whereas PCA allowed some clustering with respect to fat, in the case of the protein mixtures no discrimination was found as a function of emulsifier type. Classical methods (fat globule size distribution and melting time) were also used for ice cream characterization. Some formulations exhibited specific behaviours that made the understanding of their physical and molecular properties difficult when analysing only the physicochemical characteristics of the ingredients.