This paper proposes an engineering approach at the microstructural scale to the problem of fatigue crack initiation. Initiation of an engineering fatigue crack approximately 0.2–0.4 mm in size is broken down into the separate processes of nucleation of a small facet, and subsequent short crack growth of the facet to an engineering crack. Engineering fatigue models frequently don't consider these processes separately and have difficulty predicting the total initiation life as the distributions of the separate processes are not known. A fundamental approach to modelling initiation life that tries to address these issues has been developed. A combination of crystal plasticity and the extended finite element method are used to model nucleation followed by short crack growth. The crack growth rate is calculated using the local stored energy at the crack tip, crack growth occurs on the crystallographic plane with the largest accumulated slip, thus allowing crack tortuosity to be modelled. A grain boundary hardening model has been developed which reduces the local stored energy available for growth at grain boundaries resulting in growth retardation. The levels of retardation experienced are in agreement with published data for a range of materials. The method has been successfully correlated with fatigue data for coarse grain RR1000.