Most RBS data analysis assumes straight trajectories and one single large angle backscattering event for the analysing beam, and flat sample surface and interfaces. Multiple scattering, plural scattering, and roughness are often ignored. This may lead to erroneous data analysis. In complex systems, understanding and taking into account these effects is essential. We discuss pitfalls that can occur when they are ignored, with emphasis on multilayer systems. (c) 2005 Elsevier B.V. All rights reserved.