Reflective actions in collaborative design can potentially improve design performance and results. This paper quantitatively reexamines the relationships between reflective activities and design performance during the collaborative design process in terms of reflection in action. Twenty sets of protocol data were encoded by a modified version of Valkenburg and Dorst's coding scheme. Using statistical testing, the relationship between the design performance and the number of activities plus the transitions was examined. A significant statistical correlation was found between the percentage of mature framing (setting up of a desired goal with sufficient follow-ups) and the overall performance. These quantitative results verify the qualitative findings of the previous study.