Closed-loop identification plays an important role in improving existing underperforming closed-loop controllers, also enabling the identification of systems that, either by technical or economical reasons, cannot operate in open-loop fashion. A simple and noise-immune closed-loop identification approach based on step responses is presented. This method is entirely conceived in the time domain, and it is flexible enough to identify a set of simple low-order models that can be obtained from either reference or disturbance closed-loop step responses. This latter feature makes the proposed approach an ideal option as a building block of an adaptive controller. A comparative study considering other closed-loop identification methods available in the literature highlights the effectiveness of the new approach. Finally, a set of experimental examples demonstrates its ability to deal with real industrial processes.