Many researchers have analyzed visual language design using Cognitive Dimensions (CDs), but some have reinterpreted the purpose, vocabulary, and use of CDs, potentially creating confusion. In particular, those who have used CDs to convince themselves or others that their language is usable have tended to ignore or downplay the tradeoffs inherent in design, resulting in evaluations that provide few insights. Researchers who do not consider who, when, and how best to analyze a visual language using CDs are likely to miss the most useful opportunities to uncover problems in their visual languages. In this paper, we consider common breakdowns when using CDs in analysis. Then, using three case studies, we demonstrate how the who, when, and how circumstances under which CDs are applied impact the gains that can be expected.