Noncrimp stitched fabrics (NCFs) are often used as reinforcing materials in high-performance composite materials. Prediction models of the processing stage of the manufacturing are highly desirable in order to enhance the control of the process and enable the production of materials with higher quality. In NCFs, layers of parallel fiber bundles consisting of a large number of fibers are stitched together with other layers to form a network of interbundle channels in different directions. In earlier works, numerical simulations on unit cells had been performed in order to predict the global permeability of NCFs. It was shown that features like the thread influence the local permeability of the unit cells and therefore, the local permeability distribution of a fabric also. Furthermore, this influences the global permeability of the entire fabric. In the present paper, different geometrical features are therefore studied in order to investigate their influence on the local permeability within an NCF. The stitching process in addition to the interbundle channels, gives rise to two geometrical features, the thread which penetrates the channels and the crossing of fibers between two neighboring fiber bundles. The influences of these two features on the local permeability are studied together with variations of other geometrical parameters of the fabric. Computational Fluid Dynamics are used for the flow simulations in order to calculate the local permeability for the different unit cells. To ensure quality and trust, the numerical accuracy of the simulations is also studied. This work proves that the thread and the crossings, as well as the variations of the width and the height of the interbundle channels, have great influence on the local permeability. Prediction models therefore, have to take these features as well as geometry distortions, which influence the local permeability distribution, into account in order to make accurate predictions of the global permeability of a fabric.