Designed experiment over the model's controls. Requires at least two controllable factors to be meaningful.