You have errors in you measurement, mostly in assuming employees only add a positive ROI when in fact their value can go negative. Thus doing the bare minimum is worth more than you think as long as the employee stays in the positive ROI (which I assume is the measure of bare minimum).