The way an invasion progresses through space is a theme of interest common to invasion ecology and biological pest control. Models and mark-release studies of arthropods have been used extensively to extend and inform invasion processes of establishment and spread. However, the extremely common single-scale approach of monitoring initial spread leads to misinterpretation of rate and mode. Using the intentional release of a novel biological control agent (a parasitic hymenoptera, Eretmocerus hayati Zolnerowich & Rose (Hymenoptera: Aphelinidae), we studied its initial dispersal and spread at three different spatial scales, the local scale (tens of metres), field scale (hundreds of metres) and landscape scale (kilometres) around the release point. We fit models to each observed spread pattern at each spatial scale. We show that E. hayati exhibits stratified dispersal; moving further, faster and by a different mechanism than would have been concluded with a single local-scale post-release sampling design. In fact, interpretation of each scale independent of other scales gave three different models of dispersal, and three different impressions of the dominant dispersal mechanisms. Our findings demonstrate that using a single-scale approach may lead to quite erroneous conclusions, hence the necessity of using a multiple-scale hierarchical sampling design for inferring spread and the dominant dispersal mechanism of either human intended or unintended invasions.