How motile bacteria navigate environmental chemical gradients has implications ranging from health to climate science, but the underlying behavioral mechanisms are unknown for most species. The well-studied navigation strategy of Escherichia coli forms a powerful paradigm that is widely assumed to translate to other bacterial species. This assumption is rarely tested because of a lack of techniques capable of bridging scales from individual navigation behavior to the resulting population-level chemotactic performance. Here, we present such a multiscale 3D chemotaxis assay by combining high-throughput 3D bacterial tracking with microfluidically created chemical gradients. Large datasets of 3D trajectories yield the statistical power required to assess chemotactic performance at the population level, while simultaneously resolving the underlying 3D navigation behavior for every individual. We demonstrate that surface effects confound typical 2D chemotaxis assays, and reveal that, contrary to previous reports, Caulobacter crescentus breaks with the E. coli paradigm.