This paper presents a powerful approach to evaluating public technologies by capturing and analysing pedestrian traffic using computer vision. This approach is highly flexible and scales better than traditional ethnographic techniques often used to evaluate technology in public spaces. This technique can be used to evaluate a wide variety of public installations and the data collected complements existing approaches. Our technique allows behavioural analysis of both interacting users and non-interacting passers-by. This gives us the tools to understand how technology changes public spaces, how passers-by approach or avoid public technologies, and how different interaction styles work in public spaces. In the paper, we apply this technique to two large public displays and a street performance. The results demonstrate how metrics such as walking speed and proximity can be used for analysis, and how this can be used to capture disruption to pedestrian traffic and passer-by approach patterns.