More in practice, imagine you want to derive an implicit-solvent coarse-grained potential for a macromolecule such as a simple polymer. You decide (for instance) to use a repeating unit, containing, say, 20 atoms, as the coarse-graining unit. You will need to simulate a short-ish polymer to obtain your coarse-grained interaction (say, 20 repeating units), which will be simulated in an explicit solvent. The whole simulation might contain thousands of atoms, and you use it to parametrise the coarse-grained interaction. Once this is ready, you will be able to simulate the same system at a fraction of the computational cost, since your thousands atoms have become $$\approx 20$$ beads.