The highly heterogeneous workloads generated by current scientific and technological research complicate job scheduling in supercomputers, making commercial scheduling tools challenging to be configured effectively. Recent studies have analysed these workloads, but the impact of job scheduling policy changes on both job input and system output variables remains largely unexplored. This paper aims at a methodology for characterising both users’ and system’s behaviour in response to policy changes. As a case study, we use the Santos Dumont supercomputer at the National Laboratory for Scientific Computing (LNCC) in Brazil; we examine its job accounting records over two years, with a policy change in between them targeting a reduction in waiting times. Key findings of our study include differential impacts of the policy change on jobs with varying waiting times and the utility of grouping users to interpret eventual behavioral changes due to the policy change.