A significant proportion of energy consumption by buildings worldwide, estimated to ca. 40%, has yielded a high importance to studying buildings’ performance. Performance Testing is a mean by which buildings can be continuously commissioned to ensure that they operate as designed. Historically, setup of Performance Tests has been manual and labor-intensive and has required intimate knowledge of buildings’ complexity and systems. The emergence of the concept of smart buildings has provided an opportunity to overcome this restriction. In this paper, we propose a framework for automated Performance Testing of smart buildings that utilizes metadata models. The approach features automatic detection of applicable Performance Tests using metadata queries and their corresponding instantiation, as well as continuous commissioning based on metadata. The presented approach has been implemented and tested on a case study building at a university campus in Denmark.