A Generative AI Engineer has developed a RAG application to help employees interpret HR documentation. The prototype is functional and has received positive initial feedback from internal testers. How should the engineer now formally evaluate the system's performance and identify areas for improvement?