It is widely recognized that pollinators vary in their effectiveness in pollination mutualisms, due both to differences in flower-pollinator morphological fit as well as pollinator behaviour. However, pollination webs typically treat all interactions as equal, and we contend that this method may provide misleading results. Using empirical and theoretical data, we present the case study of a self-incompatible herb in which the number of flowers visited by a pollinator cannot be used as a surrogate for the total effect of a pollinator on a plant due to differences in per-visit effectiveness at producing seeds. In self-incompatible species, the relationship between interaction frequency and per-interaction effect may become increasingly negative as more flowers per plant are visited due to geitonogamous pollen transfer. We found that pollinators making longer bouts (i.e. visiting more flowers per plant visit) had an overall higher pollination success per bout. However, per-interaction effects tended to decrease as the bout progressed, particularly for pollinators that cause higher pollen deposition. Since the same interaction frequency may result from different combinations of number of bouts (plant visits) and bout length (flowers visited/bout), pollinators making repeatedly shorter bouts may contribute more to plant reproduction for the same number of flowers visited. Consequently, the magnitude of the differences in number of interactions of different insect types may be overridden by the magnitude of the differences in effectiveness as pollinators, even if the same pollinators consistently interact more frequently. We discuss two predictions regarding the validity of using interaction frequency as a surrogate for plant seed production (as a measure of total effect), depending on the degree of self-compatibility, plant size and floral display. We suggest that the role of interaction frequency must be tested for different species, environments, and across wider scales to validate its use as a surrogate for total effect in plant-pollinator networks.