To enhance the reliability of software, it is important is to review artifacts such as design documents to remove defects earlier. There are various review methods, and project managers face the challenge of choosing a feasible method for their current projects. One of approaches to support the selection of review methods is to evaluate review methods beforehand, to identify the most effective method on average. However, past studies have not evaluated review methods thoroughly, because the evaluation is often time-consuming. We propose a new bandit-algorithm (BA) based method which evaluates candidates of review methods during review activity, and dynamically select the best one based on the evaluation. In the experiment, we assume that the proposed method is applied to design document review on basic design phase. We performed experiments based on the simulation, instead of using actual dataset. On our simulation, when review method is selected by our method, productivity (i.e., total development time) was improved by about 1.25 times, and it was the second highest among candidates of review methods.