Ecological restoration (ER) of terrestrial ecosystems has become widespread in past decades. However, assessing its success is complex mainly due to the diversity of objectives pursued, actions undertaken but also statistical methods for treating data. We demonstrate here that, due to the heterogeneity of collected data, the success of restoration actions can be overestimated in meta-analyses. We advocate analyzing distinctly two types of actions in ER, those aiming at increasing an ecosystem attribute (e.g. species richness of a native plant species, ER+), and those aiming at decreasing it (e.g. invasive species cover, ER-). We also suggest that only one index for assessing the success of a restoration action is not enough. We propose here to complete RR (Remaining Recovery) by a novel index informing on ‘what has been restored by comparison to what should have been recovered’: the ‘Achieved Restoration’ index (AR).