Nowadays, the importance of providing frequency stability to power systems with hydropower units is significantly increasing due to the growing proportion of volatile renewable energy sources. However, the provision of such services, as frequency containment reserve, inevitably increases the wear and tear, decreasing the performance of the hydraulic machines, especially the Kaplan ones. In view of a growing demand for frequency containment reserve, it is important to quantify the performance decline and occasionally recompute the characteristic curves of the hydraulic machine to update the control system and reduce the efficiency loss. Since this process is usually expensive, as it requires off-line tests, this paper proposes a data-driven method to estimate the performance decrease of Kaplan turbines by leveraging operational recorded data. The proposed approach is applied to compute a new CAM relation for the run-of-river power plant of Vogelgrun. The new CAM, obtained with the proposed method is finally compared with real operational data.