Monitoring is essential to validate the runtime behaviour of dynamic distributed systems. However, monitors can inform of relevant events as they occur, but by their very nature they will not report about all those events that are not happening. In service-oriented applications it would be desirable to have means to assess the thoroughness of the interactions among the services that are being monitored. In case some events or message sequences or interaction patterns have not been observed for a while, in fact, one could timely check whether this happens because something is going wrong. In this paper, we introduce the novel notion of monitoring adequacy, which is generic and can be defined on different entities. We then define two adequacy criteria for service compositions and implement a proof-of-concept adequate monitoring framework. We validate the approach on two case studies, the Travel Reservation System and the Future Market choreographies.