The importance of model composition in model-centric software development is well recognized by researchers and practitioners. However, little is known about the critical factors influencing the effort that developers invest to combine design models, detect and resolve inconsistencies in practice. This paper, therefore, reports on five industrial case studies where the model composition was used to evolve and reconcile large-scale design models. These studies aim at: (1) gathering empirical evidence about the extent of composition effort when realizing different categories of changes, and (2) identifying and analyzing their influential factors. A series of 297 evolution scenarios was performed on the target systems, leading to more than 2 million compositions of model elements. Our findings suggest that: the inconsistency resolution effort is much higher than the upfront effort to apply the composition technique and detect inconsistencies; the developer’s reputation significantly influences the resolution of conflicting changes; and the evolutions dominated by additions required less effort.