Diffusion noise represents a major constraint to successful liquid state nano-NMR spectroscopy. Using the Fisher information as a faithful measure, we theoretically calculate and experimentally show that phase sensitive protocols are superior in most experimental scenarios, as they maximize information extraction from correlations in the sample. We derive the optimal experimental parameters for quantum heterodyne detection (Qdyne) and present the most accurate statistically polarized nano-NMR Qdyne detection experiments to date, leading the way to resolve chemical shifts and J couplings at the nanoscale.