The recently introduced comparison of flow pattern distributions (CFPD) method for the identification, quantification and interpretation of anomalies in district metered areas (DMAs) or supply area flow time series relies, for practical applications, on visual identification and interpretation of features in CFPD block diagrams. This paper presents an algorithm for automated feature recognition in CFPD analyses of DMA or supply area flow data, called CuBOid, which is useful for objective selection and analysis of features and automated (pre-)screening of data. As such, it can contribute to rapid identification of new leakages, unregistered changes in valve status or network configuration, etc., in DMAs and supply areas. The method is tested on synthetic and real flow data. The obtained results show that the method performs well in synthetic tests and allows an objective identification of most anomalies in flow patterns in a real life dataset.