Business processes play an important role in organizations; however, not enough attention is given to analyzing and modeling errors in them. In this paper we study syntactic and control flow error frequencies in business processes from real industry projects. Our samples come from a number of application domains such as Banking and Capital Markets, Insurance and Healthcare, and Retail. We consider industrial business processes modeled in Business Process Modeling Notation (BPMN) and use graph-theoretic techniques and Petri net-based analyses to detect syntactic and control flow related errors respectively. We then use a set of metrics that capture different network characteristics of the models and study the empirical relations between the metrics and process errors. The major results of the empirical investigation are: (a) multiple edges to or from tasks as well as hanging nodes are the predominant forms of syntactic errors (b) syntactic errors occur frequently in Retail & Logistics domain and significantly less in the Insurance and Healthcare domain and (c) the probability of error occurrence can be modeled as a function of node-size and coefficient of connectivity; the logistic regression model correctly classified 97.6% of the cases, with a sensitivity of 100% and specificity of 86.2%.