Microspectrofluorimetry offers high sensitivity, selectivity, fast data acquisition, good spatial resolution (down to 2 μm), and the possibility of in-depth profiling. It has proved to be a powerful analytical tool in identifying dyes and lake pigments in works of art. To maximize the extraction of the information present in fluorescence emission and excitation spectra, we propose a chemometric approach to discriminate dark reds to pink colours based on brazilwood, cochineal, kermes and lac dye. These range of hues was obtained using a diverse range of medieval recipes for brazilwood, kermes and lac colourants and Winsor and Newton archive for cochineal lake pigments; the lake pigments were analyzed as colour paints (arabic-gum and glair were the medieval binders selected). Unsupervised (HCA & PCA) and supervised (SIMCA) modelling were tested, allowing to explore similarities between colourants and classify the spectral data into the different lake pigments classes. It was possible to separate the four different chromophores based on their excitation spectra or bringing together the emission and excitation spectra. The first method could also differentiate between the cochineal lake pigments, in particular between crimson lakes with different aluminates and an extender (gypsum) and between carmines with different complexing ions (aluminum and calcium).