Humans alter their environment at the landscape scale, which can harm the biota and the services it provides, such as pollination of bee-attractive crops. Pollination happens when individual bees interact physically with individual flowers, so what links landscape to the quality of pollination? I present a series of mechanistic models that express seed set in a plant population as a function of the three attributes of its pollinator fauna (species composition, abundance and ‘per capita-per visit’ pollen deposition). All three can be measured quantitatively in real systems and I illustrate this by an example of pollination-yield in two crops, brinjal and mustard, in Odisha, India. By modelling pollination services at sites along a gradient of anthropogenic landscape transformation, it was possible to estimate a critical threshold at which further loss of nearby natural vegetation would limit crop yield. The threshold will be a useful benchmark for safeguarding pollination services in future. In the Odisha example, the distal link between pollinator fauna and landscape quality was alone made by statistical correlation – and this presents a prime target for future mathematical analogy.