This paper proposes an initial quality model for model composition effort, which serves as a frame of reference to developers and researchers to plan and perform qualitative and quantitative investigations, as well as replicate and reproduce empirical studies. A series of empirical studies supports the proposed quality model, including five industrial case studies, two controlled experiments, three quasi-experiments, interviews and seven observational studies. Moreover, these studies have systematically demonstrated the real benefits of using a frame of reference to enable learning about model composition effort from experimentation.