AbstractThe architecture (geometry, fault network, and stacking pattern of accreted thrust sheets) of accretionary wedges influences subduction zone processes. However, it remains challenging to constrain the architectural evolution in natural accretionary wedges over geological timescales. In this study, we undertook sandbox analog modeling, with quantitative analysis of the wedge geometry and digital image correlation‐based kinematics, to delineate the wedge growth history with four décollement settings (single or double and continuous or discontinuous). The results show that the wedge is formed by repeated episodic frontal accretion with a constant cycle (i.e., the accretion cycle), and the degree of coupling between the base of the wedge and subducting plate interface appears to depend on the relative strengths of the wedge and detachment. An interbedded décollement layer in the incoming sediment facilitated wedge segmentation and rearrangement of the internal fault network, which weakened the wedge strength. A combination of a detachable high‐friction patch in the basal décollement and a continuous interbedded weak layer enabled underplating of underthrusted sediment beneath the inner wedge, which involved a low‐angle, long‐lived forethrust and multiple cycles of frontal accretion on short‐lived forethrusts at the deformation front. Our findings suggest that décollement configuration is a key factor in controlling the accretion cycle, strain distribution, fault network, and wedge strength on timescales of ∼105 yr in natural accretionary systems. This result should be considered when investigating modern subduction zones.