It tests how sensitive the model is to the value of each individual parameter.
It does not allow us to capture parameter interactions: how the model’s sensitivity to one parameter might change as other parameters change.