How do you determine whether a change in campaign performance is statistically significant or just random noise?
