This paper presents a rigorous statistical approach to identify the controlling factors in the development of ground movements associated with deep excavations. It also gives the most suitable definition of support stiffness from many suggested definitions in the literature. The study is based on a newly compiled database from 389 case studies of propped and anchored excavations. Data mining techniques (e.g. principal component analysis and multi-linear regression) were used to identify significant relationships between the parameters under study and to quantify the global trends in the database. The study shows that the main factors controlling the ground movements are those related to ground conditions, confirming the conclusions of previous empirical studies. It is also shown that the definition of Addenbrooke et al. (1994) is the most suitable expression of support stiffness, therefore providing conclusive evidence for its future use.