Metal dispersion is a key concept in heterogeneous catalysis. The conventional approaches for its estimation strongly rely on chemisorption with different probe molecules. Albeit they can generally provide an 'averaged' value in a cost-effective manner, the inhomogeneity of the metal species and the complicated metal-support interactions pose formidable challenges for the accurate determination. Full metal species quantification (FMSQ) is introduced as an advanced method to depict the whole distribution of the metal species, ranging from single atoms to clusters and nanoparticles, in a practical solid catalyst. In this approach, automated analysis of massive high-angle annular dark field scanning transmission electron microscopic images is realized through algorithms specialized in combining the electron microscopy-based atom recognition statistics and deep learning-driven nanoparticle segmentation. In this Concept article, different techniques for determining the metal dispersion are discussed with their pros and cons. FMSQ is highlighted for it can circumvent the drawbacks of conventional approaches, allowing more reliable structure-performance relationships beyond the metal size.