GPS collars have become commonplace as a tool to monitor wildlife movement at fine and broad scales. These datasets comprise some of the most compelling evidence for the impacts of infrastructure on wildlife connectivity and can guide efforts to improve animal flow across human-modified landscapes. Movement data can be analyzed via diverse methods to select appropriate sites for wildlife fencing and crossing structures, but certain considerations emerge as common to many workflows. Early assessments of the species’ characteristic movement behavior and of the quality and availability of GPS and other datasets can inform which approaches will best reflect the ecological processes unfolding on the landscape. Communicating the methodological workflows then used in site selection for fencing and crossing structures improves stakeholder support and scientific rigor. Though each situation will undoubtedly require adapted approaches, we posit that many workflows should be driven by the questions: 1) How accurate and complete are the available data relative to the true movement process?; 2) Where do animals have the most opportunity to cross the roadway surface?; 3) Do they ever cross the road surface when given the opportunity?; 4) How do scale (both temporal and spatial) and habitat availability and selection (in terms of both risks and resources) influence crossing opportunities and events?; and 5) What other aspects of species behavior might inform structure design? We discuss two case studies where three major highways (I-15, I-40, and I-8) limit bighorn sheep connectivity in desert regions of Southern California, USA. Each barrier differs in permeability, which affected the data available to assess ideal crossing structure locations, but data included: GPS collar-recorded movements with different road crossing frequency, landscape resistance models and crossing events based on genetics, and roadkill events. Through these case studies, we discuss how the question-driven workflow was applied and how data collection and model approaches were adapted to address the circumstances of the species, data, and scale of the ecological process. We show draft and final optimal locations for wildlife fencing and crossings based on movement data. We also provide in situ over-crossing designs that provide habitat continuity, are designed for bighorn sheep, and meet bridge engineering requirements. Ultimately, most datasets only capture part of the complex interactions between animals and roads, but by recognizing the limitations of wildlife data, combining data types, considering different scales of movement and behavior, and validating predictions whenever possible, we can design mitigations that are appropriate and effective.