Accurate measurement of the convective heat transfer coefficient h(c) at external surfaces, e.g. at building facades and roofs, is of fundamental importance for heat transfer studies of the built environment. There are two basic methods for measurement of h(c), the Loveday and Ito methods, which use one and two heated sensor units, respectively. To guide in selection of method and operating conditions, and in design of the sensor, we performed an error analysis. This included estimation of systematic errors, comparison between methods, and to established Nusselt number correlations, sensitivity analysis, and an evaluation of the measurement uncertainty. The main conclusion was that both methods, at forced convection, yielded measurement uncertainties at the 4% level, provided that the Ito method was operated under the new condition, where one of its sensors remained unheated. However, at natural convection conditions, the Ito method cannot be operated with one of its sensors unheated, since h(c) is then zero at that sensor surface, which violates the method assumption that he is the same at both sensors. Sensitivity analysis showed that systematic errors will be reduced by decreasing the sensor surface emissivity. The major source of measurement uncertainty was the conductive heat flux estimate. (C) 2016 Elsevier B.V. All rights reserved.