Computing an average curve and response corridors is critical for assessing biomechanical data and comparisons to other datasets and numerical models. However, extant methods are often case-specific and lack a strong statistical foundation. A general methodology using arc-length re-parameterisation and non-linear signal registration has been proposed to provide a feature-based assessment of average biomechanical responses and statistical variability with the key advantage of the applicability of a single method to a wide range of physical responses. In this study, the arc-length-based methodology was applied to two experimental datasets:​ the compressive behaviour of porcine brain tissue and the load-unload response of a human thorax. In both cases, the arc-length corridor method captured the underlying shape of material or subject responses without a priori assumptions of response behaviour, suitable to a wide range of biomechanical data from monotonic signals without a common termination point to highly variable, hysteretic responses, and did not distort the underlying shape or variability of the average response like common contemporary methods. The arc-length corridor method is distributed freely in the software package ARCGen, available for MATLAB and Python under a permissive, opensource license (https:​//github.com/IMMC-UWaterloo).