Food supplementation experiments have been widely used to get detailed insight into how food supply contributes to the reproductive performance of wild animals. Surprisingly, even though food seldom is distributed evenly in space, variation in local habitat quality has usually not been controlled for in food supplementation studies. With results from a two-year feeding experiment involving a habitat-sensitive avian top predator, the Northern Goshawk Accipiter gentilis, we show that treatment effects on goshawk reproductive performance are habitat dependent. Extra food reduced nestling mortality in low-quality territories where prime habitat (forest) is scarce, but not in high-quality territories where prime habitat is abundant. Consequently, brood size did not differ between treatment categories in heavily forested territories, but fledgling numbers differed between unfed and fed goshawk pairs breeding in territories where forest is scarce. However, because extra food was not superabundant, this artificial increase in offspring number induced a dramatic decrease in nestling condition in low-quality territories. Treatment effects were detected even after controlling statistically for other potentially confounding effects (year, territory identity) and strongly covaried with territory-specific abundances of the most important summer prey species. These results highlight the importance of acknowledging the effect that small-scale variation in habitat quality and availability of natural food may have on the results of food supplementation experiments. In order to assess the generality of food supplementation effects, the integration of habitat heterogeneity and variation in food abundance is thus needed, especially among species in which small-scale variation in habitat quality influences demographic patterns.