We present a computational approach to understanding users' purchase behavior through personality analytics. We model purchase behavior as interactions between personality traits, consumption preferences and product attributes, and represent such interactions as digital traces. We model all possible digital traces using a likelihood trace graph, and determine the likelihood of an edge based on the associations between adjacent pairs of personality attributes, consumption preferences, and product attributes, respectively. In our approach, users' personality traits are inferred from their written texts, and are correlated to a canonical set of consumption preferences. The consumption preferences are then mapped to product descriptions based on their semantic similarity. We demonstrate the effectiveness and usefulness of our approach through a case study on a real-world purchase data.