Subscores have the potential to provide valuable remedial information to examinees; however, they typically lack the necessary psychometric properties to be useful (Sinharay, 2010). This graphic displays four visualizations regarding subscore reporting and validity based on credentialing exam data. Figure 1a illustrates a traditional subscore profile commonly used to convey an examinees subscore performance and measurement error. These profiles can obfuscate the inferences that test developers intend users to make as they omit content domain weighting and examinees struggle to correctly interpret measurement error (Clauser & Rick, 2016; Rick & Clauser, 2016). In response to these concerns, Figure 1b reconceptualizes Figure 1a as the test blueprint overlaid with performance indicators derived by comparing an examinee's score estimate to average performance—providing the categorical finding rather than tasking the examinee to make the correct inference. For instance, Figure 1a could mislead an examinee that they should primarily study content domain E, which comprises only 9% of the exam. Whereas Figure 1b makes the relative weighting of content domains explicit and illustrates that improvements in content domain A may be the most relevant for subsequent attempts. Figures 2a and 2b can guide test developers in deciding whether subscores should be reported. Figure 2a depicts the disattenuated correlations among all subscores, which in our example highlights the lack of uniqueness among the subscores (hence the overlapping profiles observed in Figure 1a). Figure 2b plots the subscores against criteria relevant in Sinharay and Haberman's value-added-ratio statistic (Feinberg & Jurich, 2017; Haberman, 2008), where “Value-Added,” “Redundant,” and “Misleading” refer to subscores that are better, similar to, or worse than the total score as a predictor of their respective true subscores. Given that the particular subscores in this example are either too unreliable or too highly correlated with the total score to provide meaningful information, it follows that any deviations observed for a particular profile are most likely due to error. Thus, in situations where subscores lack the appropriate psychometric properties, but must be reported for policy reasons, providing categorical performance feedback and test blueprint information, such as in Figure 1b, may minimize misinterpretation. If you are interested in learning more about this informative data visualization, contact the principal author Richard Feinberg (RFeinberg@nbme.org). We want to hear your feedback! Let us know what you think by emailing Ally Shay Thomas (thomasa11@upmc.edu).