Here we use time-resolved and steady-state optical spectroscopy on state-of-the-art low- and high-bandgap perovskite films for tandems to quantify intrinsic recombination rates and absorption coefficients. We apply these data to calculate the limiting efficiency of perovskite-silicon and all-perovskite two-terminal tandems employing currently available bandgap materials as 42.0 % and 40.8 % respectively. By including luminescence coupling between sub-cells, i.e. the re-emission of photons from the high-bandgap sub-cell and their absorption in the low-bandgap sub-cell, we reveal the stringent need for current matching is relaxed when the high-bandgap sub-cell is a luminescent perovskite compared to calculations that do not consider luminescence coupling. We show luminescence coupling becomes important in all-perovskite tandems when charge carrier trapping rates are < 10$^{6}$ s$^{-1}$ (corresponding to carrier lifetimes longer than 1 ${mu}$s at low excitation densities) in the high-bandgap sub-cell, which is lowered to 10$^{5}$ s$^{-1}$ in the better-bandgap-matched perovskite-silicon cells. We demonstrate luminescence coupling endows greater flexibility in both sub-cell thicknesses, increased tolerance to different spectral conditions and a reduction in the total thickness of light absorbing layers. To maximally exploit luminescence coupling we reveal a key design rule for luminescent perovskite-based tandems: the high-bandgap sub-cell should always have the higher short-circuit current. Importantly, this can be achieved by reducing the bandgap or increasing the thickness in the high-bandgap sub-cell with minimal reduction in efficiency, thus allowing for wider, unstable bandgap compositions (>1.7 eV) to be avoided. Finally, we experimentally visualise luminescence coupling in an all-perovskite tandem device stack through cross-section luminescence images.