Data flow characterization is a key issue not only in the synthesis of data processing systems, but also in control systems with data-dependent control constructs. Given the input data probability distributions of the system specification, major architectural decisions at high levels of abstraction rely on estimates of the data distributions through the rest of the behavioral description. The increasing complexity of the designs demands for methods that provide estimates fast, as oppossed to the slow conventional methods. The applicability of a particular interval method to the data characterization problem is introduced here. Initial results show that the method can significantly reduce the computation times required for estimation, also improving the estimation errors with respect to traditional simulatuin methods.