Behavior analysis of the enterprise processes is the natural response to evaluate, anticipate and improve performances. This allows knowning states of performance whose are partially or totally uncertain. We can predict and control coming behavior and cause of the forecasted one. In order to taking into account the various factors of controllable and not-controllable performance (factors climatic for example) and identifying the cause and effect relations, one have to respond to this governing need of process behavior. Plus, one have to consider the distinct data natures. In a recent contribution, the choice of the Bayesian networks for the process behavior modeling has been justified. The aim of this paper is to provide some mathematic rules which allow us to quantify the causal relations between the nodes of the Bayesian Network by using the data featuring the process and to dispose of one predictive mathematical tool for performance evaluation of any business process. Indeed, this paper is a logical continuity of recently presented paper that concerning the construction of the structure of a Bayesian network modeling the process's behavior without node quantification. Finally, we give a real industrial case integrating various possible configurations of an optical manufacturing process.