We prove the existence of an algorithm $A$ for computing 2-d or 3-d convex hulls that is optimal for every point set in the following sense: for every sequence $σ$ of $n$ points and for every algorithm $A'$ in a certain class $\mathcal{A}$, the running time of $A$ on input $σ$ is at most a constant factor times the maximum running time of $A'$ on the worst possible permutation of $σ$ for $A'$. We establish a stronger property: for every sequence $σ$ of points and every algorithm $A'$, the running time of $A$ on $σ$ is at most a constant factor times the average running time of $A'$ over all permutations of $σ$. We call algorithms satisfying these properties instance-optimal in the order-oblivious and random-order setting. Such instance-optimal algorithms simultaneously subsume output-sensitive algorithms and distribution-dependent average-case algorithms, and all algorithms that do not take advantage of the order of the input or that assume the input is given in a random order. The class $\mathcal{A}$ under consideration consists of all algorithms in a decision tree model where the tests involve only multilinear functions with a constant number of arguments. To establish an instance-specific lower bound, we deviate from traditional Ben-Or-style proofs and adopt a new adversary argument. For 2-d convex hulls, we prove that a version of the well known algorithm by Kirkpatrick and Seidel (1986) or Chan, Snoeyink, and Yap (1995) already attains this lower bound. For 3-d convex hulls, we propose a new algorithm. We further obtain instance-optimal results for a few other standard problems in computational geometry. Our framework also reveals connection to distribution-sensitive data structures and yields new results as a byproduct, for example, on on-line orthogonal range searching in 2-d and on-line halfspace range reporting in 2-d and 3-d.