A new experimental method is proposed to improve the force measurements during water impact experiments (e.g. water entry or wave impact tests) carried out with “pseudo-rigid” mock-ups. Despite the efforts of making the mock-up as stiff as possible, the impulsive nature of water impact loads may induce a transient response of the mock-up with a broad frequency content and a perturbation of the force measurements. Using the principles of momentum conservation, it can be shown that the load cell signal is the sum of the hydrodynamic forcing term and of an additional inertial term directly related to the vibrations of the structure. In the present paper, we suggest to estimate the inertial term using several accelerometers which record the response of the structure at different locations. Assuming that the structure response can be approximated by a set of natural modes, we show that it is possible to estimate the inertial term by a linear combination of the acceleration signals. The coefficients of the linear combination may be identified a priori by performing hammer tests, but in certain cases they can also be identified a posteriori using a segment of the signal time series. The performance of the method is demonstrated by considering two experimental test cases of increasing complexity. The first test case is a hydrofoil of constant section impacting water at constant speed which exhibits two-dimensional beam-like vibrations. The second test case is a segmented model of a vertical cylinder impacted by a breaking wave whose segments exhibit three-dimensional vibrations. The proposed method is efficient, conceptually simple and very simple to implement from a signal processing point of view, which makes it promising not only for water impact problems but also for other unstationary fluid–structure interaction problems.