The detection of individual variation in foraging behaviour within wild mammal populations requires large sample sizes and relies on the multifold re-sampling of individuals. However, limits for observational studies are posed by the rarity and nocturnal or otherwise elusive habits of many mammals. We propose that the detection of foraging variation within populations of mammals may be facilitated if conventional diet analysis from faeces is combined with DNA-based individual identification methods using “genetic fingerprinting” from faeces. We applied our approach to a coyote (Canis latrans) population, and showed how individuals may vary from one another in their diet profiles. Two main groups of coyotes were distinguished on the basis of their relative use of small mammals and “other vertebrates” as primary food sources, and these two groups were further subdivided on the basis of their relative use of “other vertebrates” and fruit as secondary food sources. We show that, unless a faecal sampling scheme is used that maximizes the number of different individuals included in a survey, individual foraging variation that is left unaccounted for may result in downwardly biased faecal diet diversity estimates. Our approach allows the re-sampling of individuals over time and space, and thus may be generally useful for the testing of optimal foraging theory hypotheses in mammals and also has conservation applications.