This paper is the first to predict and then validate the overall stress-strain curve and the damage sequence comprising matrix cracking, interface debonding and fibre fracture against X-ray Computed Tomography (CT) observations for a multidirectional laminate. Until recently, numerical modelling of multi-directional multi-ply composites required idealised continuum mechanics models or idealised unit cell approaches (or homogenisation method) that cannot reliably capture property variations and the complex sequence of damage events that occur upon tensile loading. Here a multiscale 3D image-based model is used to simulate stochastic crack growth in a double-notch (-45°/90°/+45°/0°/-45°/90°/+45°/0°)s carbon fibre reinforced polymer (CFRP) composite specimen subjected to tensile loading monitored by time-lapse X-ray CT. The data integration approach involves: (1) parallel simulations of meso-scale elements (MeEs) for each ply for which the orientation of the individual fibres has been extracted from an X-ray CT image, (2) local hierarchical coupling of the MeEs into a macro-scale mechanical model of the test piece, and (3) the use of a random variation in material properties where microstructural details are not revealed by the X-ray CT characterisation method.Cohesive interface elements (CIEs) are used at both scales to predict the accumulation of interface damage and crack growth. The fibre-level modelling captures the detailed damage sequence and crack morphology including fibre/matrix debonding, sliding, matrix cracking and fibre fracture events. The multiscale model is validated by comparison with the measured tensile loading curve and the damage evolution recorded by the X-ray CT. This approach can reduce the reliance of certification on extensive heirarchical structural testing schemes from test-piece to full-scale component.