In order to describe cooling processes, the thermal diffusivity alpha and the convective transfer coefficient h must be known. To this end, in general the analytical solution of the diffusion equation with only the first term is fitted to an experimental dataset of the temperature versus time, in which the temperature is measured in a known position. Despite this technique to describe well the major part of the cooling kinetics, this procedure contains a flaw for Fourier numbers less than 0.2 (lag factor). This article presents an algorithm based on optimal removal of experimental points to minimize errors in the determination of alpha and h. The algorithm was validated for several physical situations, and applied in cucumber cooling. The whole cooling kinetics was simulated with success (chi-square = 2.8756 x 10(-3) and determination coefficient = 0.9991), including the region responsible by the lag factor. (c) 2012 Elsevier Ltd. All rights reserved.