Shock buffet is a flow unsteadiness that takes place under certain combinations of angle of attack (AOA) and transonic Mach number (𝑀_∞) on the surface of a wing due to the interaction between an oscillating shock wave and a pulsating separated boundary layer. The scope of the present work is twofold. First, it consists in providing the scientific community with a better understanding of shock buffet development, with a particular focus on the role of shock motion inversion. Second, the spanwise variation of the shock oscillations is analyzed for incipient and developed buffet flows. For this reason, the flow over a wing model with aspect ratio 2 and based on the OAT15A supercritical profile is experimentally investigated at the Trisonic wind tunnel Munich (TWM) for a fixed Reynolds number (𝑅𝑒_c) of 3 × 10⁶ and numerous aerodynamic conditions within the ranges 2.5° < AOA < 6.5° and 0.71 < 𝑀_∞ < 0.78. The shock front’s vertical and spanwise shapes are reconstructed by employing the optical technique background-oriented schlieren (BOS) from the side and the top of the test section, respectively. The analysis of the shock front from the side allows for reconstructing the key states of the flow development from steady shock to developed shock buffet. Developed buffet flows are only present for time-averaged shock positions downstream of the location of maximum curvature of the wing. In the case of constant 𝑀_∞ and increasing AOA, the inversion of shock motion appears to be a necessary condition for buffet onset. Based on this finding, the hypothesis is advanced that shock buffet derives from an attempt to increase the trailing edge pressure and satisfy the compatibility condition through a reversed shock motion. However, this attempt fails due to an unstable shock-wave/boundary-layer interaction, which further reduces the trailing edge pressure and causes the shock to keep traveling upstream. Once the shock becomes weak enough, the boundary layer reattaches and the trailing edge pressure becomes too high. The downstream shock motion of the buffet cycle is then interpreted as an attempt to decrease the trailing edge pressure and satisfy the compatibility condition. However, this attempt fails due to the occurrence of boundary layer separation and a new buffet cycle begins. The careful observation of the shock front from the top gives interesting insights into the spanwise variation of the shock oscillations. The side-wall boundary layer has two main effects on the spanwise distribution of the shock statistics. First, the shock front is not completely straight. Second, the amplitude of the shock oscillations soars moving from the mid-span to the span ends. However, once shock buffet is established, the time-averaged shock front becomes straighter and displays in-phase oscillations along the span with similar amplitude. The shock oscillations seem to be the result of the superposition of several modes. Two of them, possibly linked to the side-wall boundary layer and the steady vortices in the mid-span region, cause a spanwise variation of the shock oscillations spectrum. However, the most dominant mode in the flow is the one associated with the well-known 2-D shock buffet.