Quantum physics

Why does a quantum algorithm not generally obtain every value of a function simply by placing all possible inputs into a superposition and measuring the result?