As an effective tool for data representation and processing, granular computing has been incorporated into formal decision contexts for finding granular reducts to achieve the task of mining granular rules. However, the classification performance of granular rules has not been evaluated, and this type of method is not suitable for dynamic data. To solve this problem, the current study updates granular reducts and evaluates the obtained granular rules in terms of classification performance. Concretely, we first give a theoretical analysis of updating granular reducts and granular rules and then present a novel dynamic rule-based classification model (DRCM) based on the updating mechanism. Finally, we discuss the feasibility of the proposed model and compare it with several popular classification algorithms. The conducted experiments demonstrate that the granular reducts can improve the classification ability to a certain extent and that DRCM can achieve better classification performance on some consistent datasets.