How would you test whether a new operational workflow caused a real improvement versus random variation at Chime?
