Risk assessments of alien species are usually conducted at species level, assuming that all individuals of a given species pose similar risks. However, this may not be the case if there is substantial within-species variation that could influence invasion success. We used a seed addition experiment, comprising 25 taxonomically stratified varieties of three Brassica species introduced to roadside habitats in Canterbury, New Zealand, to quantify variation in performance among species, subspecies and varieties. We aimed to assess if species was the most appropriate taxonomic level at which to evaluate invasion risk. Differences among varieties within species explained approximately 30 times more of the variation in performance (number of individuals/quadrat) than differences among species. Some of the variation among varieties was attributable to differences in seed viability. Nevertheless, differences among taxonomic groups explained only 7% of the total variation in performance; 28% was attributable to differences among plots, reflecting broad-scale environmental variation, while 65% was attributable to differences among quadrats nested within plots, highlighting the importance of fine-scale variation in the availability of suitable microsites. Policy Implications. Our seed addition experiment quantified variation in performance of 25 taxonomically stratified Brassica taxa introduced to roadside habitats. Varieties (nested within species) differed in performance far more than did species. This suggests risk assessments carried out at species level may overlook important subspecific variation in invasion risk. This is particularly true for conventionally bred and genetically modified species, which may contain taxa posing risks different to that at which the species is assessed. Consideration should be given to subjecting unassessed subspecies and varieties of plants to risk assessments similar to those applied to species.