Limited resolution and quality of seismic data and time requirements for seismic interpretation can prevent a precise description of the connections between faults. We have focused on the impact of the uncertainties related to the connectivity of en echelon fault arrays on fluid flow simulations. We used a set of 100 different stochastic models of the same en echelon fault array. These fault array models varied in the number of relay zones, relative position of fault segments, size of overlap zones, and number of relay faults. We automatically generated a flow model from each fault array model in four main steps: (1) stochastic computation of relay fault throw, (2) horizon building, (3) generation of a flow simulation grid, and (4) definition of the static and dynamic parameters. Flow simulations performed on these stochastic fault models with deterministic petrophysical parameters entailed significant variability of reservoir behavior, which cannot always discriminate between the types of fault segmentation. We observed that the simplest interpretation consisting of one fault yielded significantly biased water cut forecasts at production wells. This highlighted the importance of integrating fault connectivity uncertainty in reservoir behavior studies.