Motivated by the recent discovery of the binary neutron-star (BNS) merger GW170817, we determine the optimal observational setup for detecting and characterizing radio counterparts of nearby ($d_Lsim40$,Mpc) BNS mergers. We simulate GW170817-like radio transients, and radio afterglows generated by fast jets with isotropic energy $E_{rm iso}sim 10^{50}$,erg, expanding in a low-density interstellar medium (ISM; $n_{rm ISM}=10^{-4}-10^{-2}$,cm$^{-3}$), observed from different viewing angles (from slightly off-axis to largely off-axis). We then determine the optimal timing of GHz radio observations following the precise localization of the BNS radio counterpart candidate, assuming a sensitivity comparable to that of the Karl G. Jansky Very Large Array. The optimization is done so as to ensure that properties such as viewing angle and circumstellar density can be correctly reconstructed with the minimum number of observations. We show that radio is the optimal band to explore the fastest ejecta from BNSs in low-density ISM, since the optical emission is likelyto be dominated by the so-called `kilonova component, while X-rays from the jet are detectable only for a small subset of the BNS models considered here. Finally, we discuss how future radio arrays like the next generation VLA (ngVLA) would improve the detectability of BNS mergers with physical parameters similar to the ones here explored.