As such, LPRs offer new opportunities to harness light–matter interactions at the nanoscale, while generating renewed interest in the self-assembly of colloidal metal nanoparticles, as a scalable approach to the preparation of such plasmonic arrays. Templated self-assembly emerged as one of the most versatile approaches, being compatible with soft-lithographic techniques such as nanoimprint lithography and amenable to a variety of materials, colloids, and solvents. Templated self-assembly additionally allows the preparation of arrays where the repeating units are composed of multiple self-assembled nanoparticles (i.e., plasmonic clusters). In this system, near-field coupling can be finely tuned, thereby showing promising results in biosensing, catalysis, or plasmonic heating. In this Account, we review the preparation of ordered arrays of clusters of plasmonic nanoparticles. We present various aspects involved in the templated self-assembly of colloidal nanoparticles, with the aim of achieving at the same time close-packed structures within each cavity of the template, and uniform deposition over a large area. We then analyze the optical properties of the prepared substrates. The preparation of hierarchical structures and the possibility of tuning both the internal structure of the cluster and their organization into arrays with different lattice parameters enable control over both near-field and far-field plasmonic coupling. This unique feature of such substrates makes it possible to exploit the interplay between these two types of coupling, for the preparation of versatile functional substrates, expanding the possibilities for the integration of plasmonic arrays into functional devices for various applications. A well-established example is their use for surface-enhanced Raman scattering. On the other hand, optimization of far-field coupling provides access to plasmonic cavities for lasing or refractive index sensing. Despite two decades of fervid scientific research, the preparation and engineering of plasmonic arrays remains a relevant topic, and many directions remain largely unexplored. We conclude with a collection of perspectives and challenges that we find particularly stimulating toward future developments of the field.