Associations between personality traits and life outcomes are usually studied using the Big Five domains and, occasionally, their facets. But recent research suggests these associations may be driven by the items (reflecting personality nuances) chosen to measure these traits. Using a large dataset (N = 6126), we examined associations with 53 self-reported outcomes using domains, facets and items (markers for nuances), training and validating models in different sample partitions. Facets better predicted outcomes than domains (on average, 18.0% versus 16.6% of variance explained), but items provided the most accurate predictions (on average 20.9%). Removing domain and facet variance from items had no effect on their predictive validity, suggesting that outcome-related information was often in items' unique variances (i.e., nuance-specific). Item-based prediction also showed the highest discriminant validity. These observations, replicating previous findings, suggest that personality traits' valid associations with outcomes are often driven by narrow personality nuances.