The leading account of intrinsicality over the last thirty years has arguably been David Lewis's account in terms of perfect naturalness. Lewis's account, however, has three serious problems: i) it cannot allow necessarily coextensive properties to differ in whether they are intrinsic; ii) it falsely classifies non-qualitative properties like being Obama as non-intrinsic; and iii) it is incompatible with a number of metaphysical theories that posit irreducibly non-categorical properties. I argue that, as a result of these problems, Lewis's account should be rejected and replaced with an alternative account, which also analyses intrinsicality in terms of perfect naturalness, but which avoids these problems.