The equivalent sand grain height principle is often used for modelling the effect of wall roughness in Reynolds-Averaged Navier-Stokes (RANS) simulations on industrial applications because it implies no additional cost compared to smooth-wall simulations and it is easyto-use. This study aims at assessing the validity of such models for rough surfaces that are representative of industrial processes. Velocity profiles extracted from RANS simulations are validated against Particle Image Velocimetry (PIV) measurements on a flat plate configuration. It is shown that RANS models based on the principle of equivalent sand grain height are valid for real products' roughness. However, the arithmetic averaged height does not appear to be linearly linked to the equivalent sand grain height ℎ. Additional numerical applications on compressor test case highlight the important impact of such a hypothesis on the estimation of the performances of turbomachinery components.