Fundamental differences in the prediction of reaction rates with intermediate and heavy target nuclei compared to the ones with light nuclei are discussed, with special emphasis on stellar modifications of the rates. Ground and excited state contributions to the stellar rates are quantified, deriving a linear weighting of excited state contributions despite of a Boltzmann population of the nuclear states. A Coulomb suppression effect of the excited state contributions is identified, acting against the usual Q-value rule in some reactions. The proper inclusion of experimental data in revised stellar rates is shown, containing revised uncertainties. An application to the s-process shows that the actual uncertainties in the neutron capture rates are larger than would be expected from the experimental errors alone. Sensitivities of reaction rates and cross sections are defined and their application in reaction studies is discussed. The conclusion provides a guide to experiment as well as theory on how to best improve the rates used in astrophysical simulations and how to assess their uncertainties.