Classifying individuals based on personality and other characteristics is a common statistical approach used in marketing, medicine, and social sciences. This approach has several advantages: it improves the simplicity of data, helps data-driven decision-making, and guides intervention strategies such as personalized care. On the other hand, continuous variables are often used to classify individuals, meaning that dimensional information is reduced to several discrete classes (of individuals) and thus much information is lost through this process. Although the loss of information may be practically or pragmatically acceptable, how much information is lost and what influence this decision has on predicting external outcomes has not been systematically investigated. Therefore, in this study, we examined the predictive performance of the classification approach compared with the dimensional approach by analyzing survey data obtained from approximately 20,000 individuals concerning physical activity and psychological traits, including the Big Five personality traits. First, we classified individuals based on the dimensional data of their psychological traits and obtained several different cluster solutions. Second, these clusters were used to predict the levels of physical activity (i.e., the classification approach), which were then compared with the predictions made by the raw dimensional scales of psychological traits (i.e., the dimensional approach). The results showed that the four-cluster solution, which was supported by the standard criterion for determining the number of clusters, achieved no more than 60% explanatory power of the dimensional approach. To achieve a comparable prediction accuracy, the number of clusters must be increased to at least 20. These findings imply that the cluster solution suggested by the conventional statistical criteria may not be optimal when clusters are used to predict external outcomes.