At least they acknowledge this: "However, the table also shows that for all outcomes (with the exception of the Build Success Rate), the standard deviation exc

At least they acknowledge this: "However, the table also shows that for all outcomes (with the exception of the Build Success Rate), the standard deviation exceeds the pre-treatment mean, and sometimes by a lot. This high variability will limit our power in our experimental regressions below." What I find even stranger is that the values in the "control" and "treatment" columns are so similar. That would be highly unlikely given the extreme variability, no?