<p>The rapid advancement of in situ monitoring technology has heightened the demand for accurate real-time prediction of landslide’s time of failure (TOF) using displacement data. Yet, the complexity and variability of in situ landslide displacement data pose significant challenges, necessitating further enhancements in the generalizability and accuracy of existing prediction methods. To address this, a new data-driven approach for dynamic landslide life expectancy prediction based on the kinematic features of landslides was proposed in this study. The commonality of overall trend features and abrupt change features of kinematic data in deformation patterns that stay stable and approaching failure were investigated, based on which, the overall trend fitting test (OTFT) for preliminary failure identification was defined. Furthermore, a multistep conceptual framework integrating preliminary stability assessment, onset of acceleration (OOA) identification, and TOF prediction was proposed for dynamic real-time TOF forecasting of landslides. Case study and generalizability assessment demonstrate that the OTFT can consistently identify data with significantly increasing trends before failure, which makes it suitable for preliminary imminent landslide failure detection. The combination of change point detection methods with the deformation standard anomaly index (DSAI) enables accurate and automatic identification of landslide’s OOA, with the sliding F test demonstrating the highest accuracy. Moreover, the predicted TOF time window under the proposed conceptual framework is narrow, and the last predicted TOF that close to the actual TOF. </p>