Composite materials have been heavily used in major industrial applications like defence, automobiles and wind turbine blades for the past few decades due to their tailorability to meet specific design needs. The high-cost left-over prepreg (3–5% waste) during composite manufacturing is usually land-filled. These left-overs can be chopped into rectangular chips and, through random spatial distribution in the mould, can be compression moulded into flat, curved plates or any other complex structural part. These randomly oriented strands (ROS) reinforced transversely isotropic composite can replace conventional metals and unidirectional composite for aircraft’s secondary and tertiary loading components. In the present work, the stochastic spatial distribution of chips in compression moulded ROS reinforced composite is modelled through Representative Volume Element (RVE). The behaviour of ROS composite is analysed by performing finite element analysis. A continuum mechanics-based linear-elastic damage model is developed involving damage parameters to degrade the effective property of individual constituents of composite for constitutive material modelling through user material, UMAT in commercial finite element solver Abaqus/Standard. This work mainly focuses on failure in the constituents of the ROS composite. The four failure modes in the chip are longitudinal breakage, transverse splitting, in-plane and out-plane shear failure, and the failure mode in the matrix is isotropic. The uni-axial tensile loaded ROS reinforced composite shows transverse chip splitting and matrix failure as dominant modes of failure, which were in good agreement with the in-house experimental results.