Computational Fluid Dynamics (CFD) now plays a pivotal role in the aerodynamic design process of the majority of automotive companies. Compared to wind-tunnel experiments it offers a much quicker way to assess new designs where full-size models may either not be available or are deemed too expensive and/or time consuming to build. With new European Union regulations requiring all variants of vehicles to be tested – CFD also provides the ability to quickly add or remove components to assess their influence on fuel efficiency and performance before the final certification process. Whilst the use and trust in CFD has grown, thanks largely to the increased availability of High-Performance Computing (HPC) resources, there are still times when correlation between CFD and the wind-tunnel is less than optimum. It is for this reason that the certification process and final signoffs are still undertaken in the wind-tunnel or with coast-down tests. To improve the accuracy of CFD simulation, over the past 5-10 years there has been a shift from steady Reynolds-Averaged Navier-Stokes (RANS) based CFD methods and processes towards transient hybrid RANS-LES simulations which offer much greater accuracy for automotive geometries. In this work, we assess the accuracy of these hybrid RANS-LES simulations (Detached Eddy Simulation family in this paper) and in particular assess their sensitivity to mesh resolution, turbulence model and numerical schemes within OpenFOAM. The desired outcome being to understand the sensitivity of these factors for hybrid RANS-LES methods and to also better understand the areas needed for future research and development.