In Portland (Oregon, USA), restoration actions have been undertaken at the watershed scale (e.g. revegetation and stormwater management) to improve water quality and, where water quality and quantity are adequate at the reach scale, to increase habitat heterogeneity. Habitat enhancement in urban streams can be important for threatened species, but challenging, because of altered catchment hydrology and urban encroachment on floodplains and channel banks. To evaluate reach-scale restoration projects in the Tryon Creek watershed, we sampled benthic macroinvertebrates and conducted habitat quality surveys pre-project and over 4 years post-project. Species sensitive to pollution and diversity of trophic groups increased after restoration. Taxonomic diversity increased after restoration but was still low compared with reference streams. We found no significant changes in trait proportions and functional diversity. Functional diversity, proportion of shredders and semivoltine invertebrates were significantly higher in reference streams than in the restored stream reaches. We hypothesized that inputs of coarse particulate organic matter and land use at watershed scale may explain the differences in biodiversity between restored and reference stream reaches. Habitat variables did not change from pre-project to post-project, so they could not explain community changes. This may have been partly attributable to insensitivity of the visual estimate methods used but likely also reflects the importance of watershed variables on aquatic biota—suggesting watershed actions may be more effective for the ecological recovery of streams. For future projects, we recommend multihabitat benthic sampling supported by studies of channel geomorphology to better understand stream response to restoration actions. Copyright © 2014 John Wiley & Sons, Ltd.