A balance of accelerations acting on the ocean implies that the flow and the pressure field do not change with time. Geostrophy is the balance of the Coriolis acceleration and the horizontal pressure gradient acceleration normal to the flow. Geostrophy has been widely used to reconstruct the oceanic flow from the pressure field either from hydrographic measurements or from along-track altimeter measurements of sea-surface height. Measurements from the SWOT satellite provide for the first time two-dimensional sea-surface height fields that allow for the computation of its curvature and thus for the estimation of the gradient wind at the surface. The gradient wind is the balance of Coriolis acceleration, horizontal pressure gradient acceleration normal to the flow, and the steady (streamline) centrifugal acceleration. With the submesoscale resolving ocean model simulation eINALT100, that is developed within the ERC synergy project WHIRLS and uses a horizontal grid-spacing between 700 m and 1000 m, and with a comparison of drifter measurements to SWOT derived velocities, we show that usually geostrophy does a better job than gradient wind in predicting the flow’s speed. By using a flow-following framework, we attribute this surprising finding to that the flow is out of balance with relevant differences between streamline and trajectory curvature that compensate parts of the streamline centrifugal term. The imbalances are shown to counteract the momentum advection due to inertia and to reduce the actual amplitudes of lateral strain-rates and relative vorticities compared to those estimated using geostrophy and gradient wind. Furthermore, the imbalances are associated with important pressure gradients tangential to the flow, which contribute to the temporal evolution of the pressure field. Processes that contribute to the imbalances are eddy-propagation, inertial oscillations, and centrifugal leakage. The presentation highlights the usefulness and intuitiveness of the flow-following framework for investigating ocean dynamics.