The Common European Framework of Reference (CEFR), with its illustrative scales and salient features of spoken language at the reference levels, is widely used as the base for rating scales for performance testing. If practitioners want to measure and report even small gains in proficiency, they need to adapt the descriptors to their local context and make further subdivisions. However, progression from one band to the next is often defined intuitively without much empirical support. This article presents an attempt to create a finer, empirically-based differentiation for academic speaking at C1 and C2 in the form of common reference points, which specify progression in a way that is more precise than the level descriptions in the CEFR but not too specific to lose their referential nature. To validate the progression, the common reference points are compared to the results of a keyword analysis of C1 and C2 descriptors for speaking from several CEFR-related sources. The results appear to confirm the soundness of the suggested progression in general terms. The findings reflect both the potential and the limitations of a keyword analysis for present purposes, indicating that the approach taken is complementary to other forms of validation.