White light images of coronal mass ejections (CMEs) are projections on the plane‐of‐sky (POS). As a result, CME kinematics are subject to projection effects. The error in the true (deprojected) speed of CMEs is one of the main causes of uncertainty to Space Weather forecasts, since all estimates of the CME time‐of‐arrival (ToA) at a certain location within the heliosphere require, as input, the CME speed. We use single viewpoint observations for 1,037 flare‐CME events between 1996 and 2017 and propose a new approach for the correction of the CME speed assuming radial propagation from the flare site. Our method is uniquely capable to produce physically reasonable deprojected speeds across the full range of source longitudes. We bound the uncertainty in the deprojected speed estimates via limits in the true angular width of a CME based on multiview‐point observations. Our corrections range up to 1.37–2.86 for CMEs originating from the center of the disk. On average, the deprojected speeds are 12.8% greater than their POS speeds. For slow CMEs (VPOS < 400 km/s) the full ice‐cream cone model performs better while for fast and very fast CMEs (VPOS > 700 km/s) the shallow ice‐cream model gives much better results. CMEs with 691–878 km/s POS speeds have a minimum ToA mean absolute error (MAE) of 11.6 h. This method, is robust, easy to use, and has immediate applicability to Space Weather forecasting applications. Moreover, regarding the speed of CMEs, our work suggests that single viewpoint observations are generally reliable.