Today IT infrastructures run an increasing number of business processes on distributed transaction processing systems. Interdependencies between the processes make performance management a challenging task. Our analysis of real-world workload patterns shows that demand is non-stationary and volatile. We introduce an outlier detection based on autoregressive forecasting methods, which detects unusual demand peaks allowing controllers to respond early. We also use static prioritization based on the workflow’s business values, in order to control the demand volatility. We show that static prioritization may lead to an improvement of performance measures, but we also found situations where it leads to inefficiencies. We therefore introduce an approach to dynamically prioritize workflows by continuously adapting prices for service-usage of incoming workflow requests. The approach dominates static prioritization in a number of experiments.