Heterogeneous catalysts are complex by nature, making particularly difficult to assess the structure of their active sites. Such complexity is inherited in part from their mode of preparation, which typically involves coprecipitation or impregnation of metal salts in aqueous solution, and the associated complex surface chemistries. In this context, surface organometallic chemistry (SOMC) has emerged as a powerful approach to generate well-defined surface species, where the metal sites are introduced by grafting tailored molecular precursors. When combined with thermolytic molecular precursors (TMPs) that can lose their organic moieties quite readily upon thermal treatment, SOMC provides access to supported isolated metal sites with defined oxidation state and nuclearity inherited from the precursor. The resulting surface species bear unusual coordination imposed by the surface that provides them high reactivity in comparison with their molecular precursor. In addition, these molecularly defined species bare strong resemblance with the active sites of supported metal oxides. However, they typically contain a higher proportion of active sitesmaking structure-activity relationship possible. They thus constitute ideal models for this important class of industrial catalysts that are used in numerous applications such as olefin epoxidation (Shell process), olefin metathesis (triolefin process), ethylene polymerization (Phillips catalysts), or propane dehydrogenation (Catofin and related processes). This SOMC/TMP approach can thus provide detailed information about the structure of active sites in industrial catalysts, their mode of initiation and deactivation, as well as the role of the support and specific thermal treatment on the final activity of the catalysts. Nonetheless, these structurally characterized surface sites still exhibit heterogeneous environments borrowed from the support itself, that explain the intrinsic complexity of heterogeneous catalysis. Furthermore, SOMC/TMP can also be used to generate and investigate supported metal nanoparticles. Starting from the well-defined isolated sites, that also contain adjacent surface OH groups, one can graft a second metal and then generate after treatment under hydrogen small and narrowly dispersed alloys or nanoparticles with tailored interfaces that can show improved catalytic performances and are amiable todetailed structure-activity relationships. This approach is illustrated by two case studies: (1) formation of supported copper nanoparticles at tailored interfaces that contain isolated metal sites for the selective hydrogenation of carbon dioxide to methanol, allowing for a detailed understanding of the role of dopants and supports in heterogeneous catalysis, and (2) preparation of highly selective and productive propane dehydrogenation catalystsbased on silica-supported Pt xGa y alloy. Overall, this Account shows how the combination of SOMC and TMP helps to generate catalysts, particularly suited for elucidating structural characterization of active sites at a molecular-level which in turn enables structure-activity relationship to be drawn. Such detailed information obtained on well-defined catalysts can then be used to understand complex effects observed in industrial catalysts (effects of supports, additives, dopants, etc.), and to extractinformation that can then be used to improve them in a more rational way.