EFT is a recipe to study observables at a scale S, that corresponds to the way we do physics: choose appropriate variables, then perturb in (hopefully small) scale ratios. I will review how to parametrise heavy "New Physics" using EFT, then discuss two examples: the relation between Higgs decay to μe and μ→eγ, and the search for four-fermion contact interactions in dilepton final states at the LHC.