The benefits of insect pollination to crop yield are used to justify management decisions across agricultural landscapes but current methods for assessing these benefits may underestimate the importance of context. We quantify how the effects of simulated insect pollination vary between five faba bean cultivars, and to what extent this changes between years, scales, yield parameters, and experimental methods. We do this by measuring responses to standardised hand pollination treatments in controlled experiments in flight cages and in the field. Pollination treatments generally improved yield, but in some cases yield was lower with additional pollination. Pollination dependence varied with cultivar, ranging from 58% (loss in yield mass per plant without pollination) in one cultivar, to a lower yield with pollination in another (−51%). Pollination dependence also varied between flight cage and field experiments (−10 to 37% in the same cultivar and year), year (4 to 33%; same cultivar and yield parameter), and yield parameter (−4 to 46%; same cultivar and year). This variability highlights that to be robust, assessments of pollination benefits need to focus upon marketable crop outputs at a whole-plant or larger scale while including and accounting for the effects of different years, sites, methodologies and cultivars.The contribution of insect pollinators to crop yield has been used to support conservation measures 1 , provide cost-effective agronomic advice 2 , and to support agricultural policies at a national level 3 . Globally, benefits of insect pollination to food production are estimated to be worth $235-577 billion per year 1 . This value is based upon estimates of the "pollinator dependence" of each crop; the proportion of harvestable yield that depends on animal mediated pollination, or how much yield is lost in absence of such pollination.Recent studies have begun to show that the pollinator dependence of a crop is not fixed and that it interacts with other biological factors (soil quality and predation by pests 4-6 ) or agronomic inputs (fertilizer, agrochemicals and water 7-11 ), see review 12 .Pollinator dependence can also vary with crop cultivar, for example in apples 13 and oilseed rape 14 . Understanding pollinator dependence on a per-cultivar basis could enable crop producers to secure production as pollination services become less predictable 15,16 following shifts in pollinator populations through time 17,18 . In the short term, producers in landscapes with low levels of semi-natural habitat and pollination service capacity 19 may benefit from use of low-dependent cultivars, though this needs to be considered alongside other agronomic attributes (e.g. yield potential, disease resistance).Quantifying variation in pollinator dependence between cultivars has rarely been the primary aim of experiments. Without standardised experimentation it is difficult to identify those cultivars with high or low pollination dependence; as above, each study that estimates the yield benefits of pollination does so...