In the framework of sustainable forest management, measuring site quality and predicting site productivity remain a major forestry topic. Over the past years, it has been fostered by a number of site-growth modelling studies seeking to establish quantitative relationships between site index and explicit biophysical indicators. In addition, comparative modelling studies of site index and site productivity have highlighted limited accordance of their environmental determinism, questioning site index as a reliable indicator of site quality. Lastly, process-based modelling approaches have recently arisen as a means to predict site productivity over large regions. All these investigations have however drawn limited attention in recent syntheses. In this review, we provide an overview of the literature on these site-growth studies. Concepts and vocabulary related to site productivity are introduced. Taking regional studies as a baseline, we first highlight recent progress regarding the geographic areas encompassed, and the major role played by NFI programmes and spatialized environmental information. A trade-off between model accuracy and geographic extent is suggested, pointing out potential deficiencies in the modelling of site-growth associations, insufficient accuracy or resolution in climatic data, or uncontrolled factors in site-growth models that emerge on a higher spatial scale. Inappropriate use of biophysical classifications where environmental factors remain implicit is also emphasized. In a next step, we discuss early and most recent indications on weaknesses of the site index concept when applied over large regions, including its differential response to climate relative to site productivity, regional variations in site index curves, site index dependence on stand density and subsequently on regional silvicultural practices. The role of genetic structure of tree populations and its integration into site-growth studies is also reviewed and discussed. The interests, limits and recent advances of process-based models as an alternative to evaluate site productivity are considered, as they may overcome some of the previous limitations. We last draw challenges and perspectives on the issue. We suggest that the accuracy – and the need – of site index as a founding concept of forestry science is questioned, by building direct productivity–environment relationships based on NFI databases as a realistic option at hand. We formulate perspectives regarding the accuracy, resolution and enlargement of environmental indicators currently used, the inclusion of information on genetic structure of tree populations in the context of adaptation to future climate change, as well as the use of site productivity models in forest management.