This paper extends the applicability of generative design for space planning frameworks for ongoing and guided post-occupancy modifications. It involves the comparison of a graph-based productive-congestion simulation with empirical data and the use of a metaheuristic search algorithm to calibrate and fine-tune simulation parameters for greater accuracy. This methodology is demonstrated through a real-world generative designed case-study and the post-occupancy collection and processing of movement data through custom computer vision workflows.