Abductive Model-Based Diagnosis (MBD) provides an intuitive approach to fault identification by reasoning on a description of the system to be diagnosed. Nevertheless, its computational complexity hinders a vast adoption and thus motivates further evaluation of efficient methods. In this paper, we investigate the structural metrics inherent to models and diagnosis problems generated on the basis of Failure Mode Effect Analysis (FMEA). Proceeding on the metrics developed, we investigate their potential as classification features to identify the most suitable diagnosis algorithm for a particular diagnosis problem. Evaluated on artificial and practical samples, our approach shows that the classifier trained on the described metrics is able to indicate the most efficient method in case of a specific diagnosis scenario.