You don't need to bring a ruler to estimate partial derivatives the way that this problem asks you to.
Expect one about a min/max problem, something about Lagrange multipliers, something about the chain rule and something about constrained partial derivatives.
And we have learned how to package partial derivatives into a vector, the gradient vector.
It has only partial derivatives for each variable.