Cluster-C personality disorders (PDs) are prevalent and impairing conditions for which schema therapy, delivered both individually (IST) or in a group (GST), is a promising treatment. However, little is known about which patients benefit most from which treatment format. This study applied machine learning methods to examine whether psychological traits can predict treatment outcome. Data was drawn from a multicenter randomized controlled trial comparing GST, IST, and treatment as usual (TAU) in patients with cluster-C PDs. Treatment outcomes included self-esteem and self-ideal discrepancy. Predictors included treatment form, diagnosis (APD, OCPD, DPD), and four psychological traits (autism, childhood trauma, sleep quality, introversion). Analyses followed a Personalized Advantage Index framework, combining linear mixed models, penalized regression, random forests, and gradient boosting, with SHAP values used for interpretability. Contrary to expectations, neither treatment form nor diagnosis predicted outcome. Instead, dimensional traits, particularly introversion, sleep quality, and childhood trauma, emerged as the most relevant predictors, especially for self-esteem. Across models, baseline outcome scores dominated prediction, and overall explained variance was low to modest. SHAP values aligned closely with traditional feature-importance metrics. These findings support a trait-based, transdiagnostic approach to personalizing treatment.