An algorithm to investigate nonlinear systems with time delay variability is proposed. It automatically cycles through segments of an open loop experiment, capturing the directional change in the output via a pair of regression equations, and estimates the time delay accordingly. The method is a blackbox, statistical approach that does not require a dynamic model nor any knowledge of the systematic causes of the time delay variability. It is designed for systems with approximately integrating behaviour, hence it is used in this article for the analysis of hydraulically actuated robotic manipulators with this characteristic and time-varying delays. In the context of control, the new algorithm provides insight into the variable time delay behaviour and hence can guide control design decisions e.g. by uncovering state dependencies.