Scanning thermal microscopy allows thermal characterization with nanoscale resolution. However, quantitative usage has been met with skepticism, because no standard exists for calibrating probe–sample thermal exchange. In this paper, three published strategies for calibrating probe–sample thermal exchange are directly compared, then used to measure bulk and thin-film thermal conductivity. It is shown that with an appropriately calibrated probe and film-on-substrate heat conduction model, thermal conductivity values of ultrathin-film (2.9–202 nm) Al2O3 on silicon are within 20% deviation of independently measured values, while more commonly used methods yield values that may deviate by more a factor of two.