In order to optimize their revenues and profits, an increasing number of businesses organize their business activities in terms of business processes. Typically, they automate important business tasks by orchestrating a number of applications and data stores. Obviously, the performance of a business process is directly dependent on the efficiency of data access, data processing, and data management. In this paper, we propose a framework for the optimization of data processing in business processes. We introduce a set of rewrite rules that transform a business process in such a way that an improved execution with respect to data management can be achieved without changing the semantics of the original process. These rewrite rules are based on a semi-procedural process graph model that externalizes data dependencies as well as control flow dependencies of a business process. Furthermore, we present a multi-stage control strategy for the optimization process. We illustrate the benefits and opportunities of our approach through a prototype implementation. Our experimental results demonstrate that independent of the underlying database system performance gains of orders of magnitude are achievable by reasoning about data and control in a unified framework.