Entanglement is one of the most studied features of quantum mechanics and in particular quantum information. Yet its role in quantum information is still not clearly understood. Results such as (R. Josza and N. Linden, Proc. Roy. Soc. Lond. A 459, 2011 (2003)) show that entanglement is necessary, but stabilizer states and the Gottesman-Knill theorem (for example) imply that it is far from sufficient. I will discuss three aspects of entanglement. First, a quantum circuit with a "vanishingly small" amount of entanglement that admits an apparent exponential speed-up over the classical case. Second, I will discuss techniques for lower-bounding the amount of entanglement in bipartite quantum states. Finally, I will discuss the role of entanglement in quantum metrology. Specifically, I will show that entangling ancillas can make no difference to the accuracy of a quantum parameter estimation, regardless of the nature of the coupling Hamiltonian. I will conclude by discussing strategies for improving the scaling of quantum parameter estimation.