Prediction of composite fatigue is not a straightforward matter, depending on various failure modes and their interactions. Scanning the literature it is obvious that there is no ‘ready to use’ model today. It cannot be expected that uniaxial ply failure data, established for uniaxial loading on unidirectional laminates, would work on a laminate basis due to, for example, effects such as interlaminar stresses. The methodology currently suggested is based on the concept of fatigue failure functions, where the key question is how to establish the longitudinal, transverse and shear fatigue functions. In this paper the fatigue failure functions are established on a laminate level and thus the calculated uniaxial properties, in this case with the Tsai-Hill failure criterion, will reflect the overall behaviour of the laminate. This also means simultaneous failure at a chosen well-defined time with the same failure index throughout the laminate. To establish the fatigue failure functions, a documented/reported S-N curve for a multi-directional laminate is considered as a series of failure points where the failure in the short life-region is essentially assumed to be fibre fracture and in the long life-region to be controlled by matrix properties. For fibre failure it is also assumed that once essential fibre failure starts to occur the remaining time to total failure is very short. The difference between first and last ply failure, in a fatigue sense, is of less importance. Once ply failure data are established they are used in a semi-log representation to calculate the fatigue life which is a straightforward matter since normalized, calculated, ply failure data collapse into one curve. Different stress ratios ( R = σ min σ max ) are presently handled with the Goodman correction approach. For combined loading, a series of failure envelopes representing different times in a log-scale constitutes the behaviour.