"Nothing is ever perfect" - a saying which also fits data elicitation instruments. Such instruments all reveal one or other failing which serve to question the reliability and objectivity of the resulting findings. Despite the well-known and obvious solution of data triangulation (cf. Neumann 1995), such is not widely implemented in current research (cf. Kasper 1998). This article concentrates on the triangulation of learner data, and in particular on the combination of product and process data. A longitudinal study of the development of the L2 pragmatic competence of 33 Irish learners of German during two terms in the target speech community serves to illustrate the necessity of triangulation to ensure that research findings are not a mere product of the elicitation instrument. Product data are elicited using a discourse completion task (DCT) and process data using retrospective interviews. The realisations of offers and refusals of offers on the level of the exchange is the focus of interest.