Coastal shorelines are a key interface between terrestrial and aquatic ecosystems and are vital for human livelihood. As a result, shorelines have experienced substantial human modifications worldwide. Shoreline “hardening” – the construction of armor including seawalls, bulkheads, or rip-rap – is a common modification that has substantial negative ecological effects. Currently, restoration involving the removal of armor and replacement with “living” shorelines is becoming an established practice. Still, the ecological response to armor removal is oftentimes unpredictable and site-specific. We hypothesized that the confluence of larger-scale geophysical features might strongly influence ecological restoration outcomes at particular locations. To measure the effectiveness of armor removal in the context of broad-scale geophysical features across the Salish Sea, WA, USA, we studied 26 paired restored and natural reference beaches of the same shoretype (feeder bluff, accretion shoreform, or pocket beach), as well as corresponding fetch, sub-basin, and percent of shoreline sediment drift cell armored. Sites were restored for an average of six years. We gauged restoration effectiveness based on levels of five ecological response variables: beach wrack (percent, depth), logs (count, width), sediments (percent sand), vegetation (percent overhanging, count of fallen trees), and insects (density, taxa richness). We found that armor removal often restored these variables to natural levels, but that restoration response was dependent on geophysical features such as shoretype and fetch. Natural beaches did have higher measurements of overhanging vegetation, fallen trees, and insect taxa richness, as these features likely take time to mature at restored beaches. Feeder bluffs had a higher proportion of surface sand and number of fallen trees than other shoretypes, coinciding with the erosion of bluff material, whereas natural pocket beaches within bordering rocky headlands had higher insect densities. Sites with a large fetch had higher input of deposited wrack and logs, whereas sites with a small fetch had higher input from localized terrestrial sources – fallen trees and eroding sand. By incorporating the effectiveness of restoration with landscape features such as shoretype and fetch, we can more effectively plan for future restoration actions and better predict their outcomes.