There is abundant literature on the consequences of academic engagement with the industry, on overall scientists' commercialization and scientific performances. Nevertheless, the literature remains silent on how those performances are contingent on the research orientation's choice at the project level. This paper aims to fill this gap by exploring different research orientations in the context of university-industry collaborative Ph.D., a topic of interest as private companies are becoming more involved in Ph.D. training. To do so, we relied on a unique dataset comprising of 635 Collaborative Ph.D. projects through the CIFRE program in France. We classified the projects' ex-ante research directions: market-pull-oriented (MPO), research-push-oriented (RPO), and simultaneous-discovery-invention-oriented (SDI), and we observed their ex-post performances. First, as expected, an orientation towards industry needs conduct to higher commercialization performances. However, counter-intuitively, those projects are also prone to have similar scientific performances than those oriented towards scientific discoveries. Second, while SDI projects were considered over-performing other research orientations, they led to more significant scientific performances than traditional orientations but generated as many patents as MPO projects. Finally, we highlight that initial research orientation is a crucial determinant variable of scientific and commercialization performances, and our paper opens rooms for further research to the literature on academic engagement, university-industry collaborations, and Collaborative Ph.D.