Compression moulding experiments of sheet moulding compound, visual observations of a vacuum test with prepregs and numerical models with two main approaches for computational fluid dynamics simulations of the mould filling phase are presented. One assumes that there are layers near the mould surfaces with much less viscosity and the other only use one viscosity model. The numerical experiments showed that the pressure could be accurately predicted with both approaches. The property necessary to predict correct pressure with altered mould closing velocities was that the bulk material had to obey shear-thinning effects. Preheating effects before compression were neglected, but altering the heating time until the prepreg was assumed to start flow had a significant effect. The experiments confirmed that the pressure is predominantly affected by the mould closing velocity. Regardless of the considered process settings, a first pressure top always appeared approximately at the logarithmic strain 0.25. A second top was associated with a slowdown of the press. The location of this was affected by the velocity and the vacuum, the latter indicating that vacuum assistance prevents a build-up of back pressure. Furthermore, heated prepreg above a critical temperature is observed to swell immediately as vacuum assistance is applied.