Recent high precision experimental data for heavy-ion fusion reactions at sub-barrier energies systematically show that a surprisingly large surface diffuseness parameter for a Woods-Saxon potential is required in order to fit the data. We point out that experimental data for quasielastic scattering at backward angles also favor a similar large value of the surface diffuseness parameter. Consequently, a double folding approach with a short-range imaginary potential for the compound nucleus formation fails to reproduce the experimental excitation function of quasielastic scattering for the $^{16}\mathrm{O}+^{154}\mathrm{Sm}$ system at energies around the Coulomb barrier. We also show that the deviation of the ratio of the quasielastic to the Rutherford cross sections from unity at deep sub-barrier energies offers an unambiguous way to determine the value of the surface diffuseness parameter in the nucleus-nucleus potential.