Crack length measurements with high accuracy are often difficult to achieve during fatigue crack propagation testing under non-isothermal conditions. In this work a modified approach to the compliance method defined in e.g. ASTM E647 is described, which is better suited for high loads, varying temperatures and for taking the scatter in Youngs modulus into account. A numerical finite element study is performed for a single edge notch specimen, to investigate the influence of initiation locations on the accuracy of the method. The change in cracked area versus change in stiffness for three different cases are numerically shown to collapse to one curve, i.e. the result is not significantly affected by how the crack is initiated. The numerical study is compared to results from two experiments using different materials, with heat tinting during the tests for extracting snapshots of the crack fronts. A good agreement between the experiments and the numerical study is shown. A new compliance curve and a new geometry function for the stress intensity factor is proposed for the single edge notch specimen. (C) 2016 Elsevier Ltd. All rights reserved.