Identifying the most important variables that determine patterns and processes is one of the main goals in many scientific fields, including ecological and evolutionary studies. Variable or relative importance is generally seen as the proportion of the variation in a response variable explained directly and indirectly by a specific predictor. Although partial regression coefficients is perhaps the most frequently used, ‘standard’, statistical technique in ecological and evolutionary studies, beta weights are inadequate indices of variable importance when predictors are intercorrelated, which tends to be the rule in most observational data sets. Among other statistical techniques, random forests and hierarchical partitioning are designed to cope with collinearity but are still much less used than beta weights to measure variable importance. Here, we compared random forests and hierarchical partitioning with linear mixed models to attempt to unravel the individual and shared variation of environmental, economic, and human population factors with success of alien species richness in eight taxonomic groups at a global scale. Results showed that random forests and hierarchical partitioning generally agreed in ranking variable importance but showed considerably different conclusions to the standard statistical approach. Specifically, random forests and hierarchical partitioning attached more importance to economic and human population variables in explaining spatial patterns of alien species richness than did region area and mean air temperature, which were emphasized more by the standard approach. Beta weights also tended to highlight less correlated predictors, such as sampling effort and precipitation. Variable importance in random forests attached more importance to economic than population variables and to absolute rather than relative predictors. Variable-importance measures are an underused statistical technique, particularly compared to the partial regression coefficients, but can shed light on evaluating the individual and shared variation of predictors of success of biological invasions and of many other biological and scientific questions.