X-ray fluorescence (XRF) spectroscopy is one of the most fundamental and trusted analytical techniques for rapid analysis of geological samples. Geological samples are typical because of their inhomogeneous morphology and diversified matrices, unusual patterns of association of elements with soils and rocks, variability in grain size and composition. XRF spectra of a simple geological motif thus can produce very complex variables and possibilities. To handle these large quantum of data, one needs sophisticated data processing tools and leveraging the same in X-ray spectrometric experiments can indeed provide propitious results. Chemometric methods when applied to XRF can be a powerful tool to categorically differentiate samples into desired groups, to quantitatively characterize clustered, overlapping spectra and above all to provide veiled interrelation among variables. This chapter intends to delineate the multivariate approaches being used in the field of XRF analysis for geological samples. Concepts of spectral pre-processing, selection of variables, data dimensionality reduction and pattern recognition has also been addressed. Principal component analysis, hierarchical cluster analysis and partial least squares regression are the main components which seek attention. Other significant methods of chemometrics include factor analysis, linear discriminant analysis, support vector machines, artificial neural networks, multivariate cure resolution and data fusion.