We investigated how the sampling process of microhistological faeces analysis could be optimised for an accurate estimation. Spring diet composition of European hare (Lepus europaeus Pallas, 1778) was determined in a juniper shrubland at Bugac, Hungary. Both inter and intraobserver reliability was high permitting us to separate the components of variance due to the methodological steps in the faeces analysis. Estimates varied depending on the number of independent droppings, pellets/individual, subsamples/pellet and epidermis/subsample. The variance was much higher among than within the independent pellet groups. The cumulative frequency estimate stabilised at around 100 epidermis fragments per pellet. We conclude that the most critical steps of the sampling procedure are the collection of independent droppings and the identification of a sufficient number of epidermis fragments. We propose to collect at least 10 independent droppings, one pellet/individual, and analyse 100 epidermis fragments as an optimum for estimating the relative frequency of forage classes reliably. The importance of the individual variability in the diet should be emphasised.