Collaborative business process modeling is a collective activity where team members jointly discuss, design, and document business processes. During such activities, team members need to communicate with each other to coordinate the modeling activities, propose and justify changes, and negotiate common terms and definitions. Throughout this process, stakeholders should be aware of when and what kind of changes have been made by each team member on the shared space so that they can discuss design ideas and build on each other's work. Joint visual attention has a fundamental role in establishing and maintaining common ground among interlocutors in such cooperative work settings. In addition to this, the co-constructed model's quality is often considered a key evaluation outcome measure to assess the success of collaboration. However, process and outcome measures of collaboration have been prone to difficulties due to challenges in devising measures that can adequately capture the complex dynamics of collaborative work. This study explored the relationship between a popularly used outcome measure in the business process modeling literature and a process measure approximating the level of joint visual attention present among the participants based on the degree of gaze cross-recurrence among the team members over a shared task space. The results suggest that joint visual attention as operationalized in terms of gaze cross-recurrence was a strong predictor of the syntactic, semantic, and pragmatic qualities of collaboratively produced business process models. Moreover, the collaboration process was subjected to qualitative analysis to probe further into the interactional organization of the modeling activity, which identified communication, coordination, awareness, group decision making, and motivation dimensions as key factors contributing to the quality of collaboration among group members. The results indicated strong relationships between the distribution of quality factors and the degree of gaze cross-recurrence and the final models' syntactic and semantic quality scores. Given the increasing availability of affordable eye trackers and the low resolution, practical nature of the employed analysis methodology, the proposed approach can be fruitfully employed to evaluate team performance and test the effectiveness of software interfaces designed to support collaborative work.