This paper examines the suitability of employing a mixed-method approach in the study of pragmatic variation in service encounters (SEs). To this end, it examines complaint strategies in a variety of data sets. Controlled-elicitation data (discourse completion tasks, role plays) are scrutinized critically as to their usefulness for SE variation research. Also, various types of naturally occurring data (specifically telephone call centers and fly-on-the-wall documentaries) are evaluated and compared. While the value of controlled data is limited at best, it is argued that they can still provide useful input for the qualitative analysis of authentic data. Moreover, given the context-sensitive nature and variability of complaints, it is suggested that variation should also be analyzed in terms of the communication medium employed (face-to-face, telephone, computer-mediated), using authentic data.