<p>Recommendation letters are a ubiquitous yet puzzling institution in professional labor markets. While criticized as uniformly positive and uninformative signals, they remain central to hiring. This paper argues that their value lies not in their content alone but in their position within a relational network, offering a formal account of one mechanism through which relational information may mitigate information asymmetry. I adopt category theory as a modeling language to represent the job market as a category interconnecting applicants, referees, and firms through morphisms like applications and recommendations. My central claim, informed by the Yoneda Lemma, is that an applicant’s evaluative identity can be formally represented through the pattern of incoming relations from others, rather than through intrinsic attributes alone. This relational-first account explains how letters may acquire evaluative significance by composing into a structured web of endorsements that standard network analysis can miss. The framework generalises graph-based models, clarifies coherence conditions for evaluation, and formalises criteria shifts as natural transformations. I discuss limitations–including the static nature of the model–and sketch extensions to dynamic hiring processes and fairness-oriented institutional design. The aim is to provide economists with a rigorous lens for analyzing evaluation, signaling, and coordination in relational economic systems.</p>