If we had first performed the linear interpolation in the "y" -direction and then in the "x" -direction, the resulting approximation would be the same.
Its main feature is the gradient approximation that requires only two measurements of the objective function, regardless of the dimension of the optimization problem.