Trust has become of increasing interest in many scientific domains, including economics, social psychology, sociology, and marketing (Blois 1999). Marketing studies have demonstrated that trust is an essential ingredient for successful relationship marketing (Doney and Cannon 1997; Morgan and Hunt 1994). In fact, Berry states: “Relationship marketing is based on the foundations of trust” (1995, p. 242). Theorizing that successful relational exchanges are motivated by trust and commitment, relational theory implicitly assumes that transactional and weak relational exchanges are not similarly motivated. According to this assumption, Garbarino & Johnson (1999) have proposed and confirmed that in business relationships for customers with a weak relational orientation trust is a peripheral evaluation and not predictive of repurchase intent. This finding contrast with the impact of trust for customers with a strong relational orientation, where trust is a key determinant of repurchase intent. Extending the work from Garbarino & Johnson (1999) we take a different approach and investigate different facets of risk as moderators of the trust – repurchase intention relation. Furthermore building on the embeddedness notion we propose and show that close personal relationships between boundary personnel of interacting companies are a key determinant of trust. Several marketing scholars (Sheth and Parvatiyar 1995; Witkowski and Thibodeau 1999) have argued for the relevance of close interpersonal relationships specifically in a business-to-business context. Consequently building close personal relationships with key customers was proposed as a way to achieve competitive advantage (e.g., Gremler and Gwinner 2000; Reichheld 1993). However our results imply that the strategy to build close personal relationships with key customers is not effective in conditions in which trust is not important; notably in situations with low risk. Our notion of risk as a moderator for the trust-loyalty relation can also explain the negligible relevance of personal relationships in Wathne, Biong and Heide's (2001) study, where risk was essentially zero.