When people make interventions to a system they expect the effects to be nearly instantaneous. Unfortunately, in most of the cases the intervention intended to improve the process actually causes outcomes to get worse before they get better, if they get better at all. The challenge in these types of situations is being able to readjust expectations that there is a delay in the improvement. This is not simply a case of learning curve where people get better by performing repetitive tasks over time. What we are describing is a delay in the improvement and, in some cases, a degradation of performance. In this paper we discuss why humans tend to underestimate such delays in process improvement across a variety of circumstances. To illustrate this, we compare data collected from a survey with three well-documented scenarios of process improvement: the implementation of a preventative maintenance program at DuPont, the modification of Tiger Woods' golf swing, and the implementation of a platform engineering initiative for the embedded software product line at Rolls-Royce. We discuss potential reasons for the chronic underestimation of these types of improvements and recommend mechanisms for making these estimates more realistic.