Higher-order languages that encourage currying are typically implemented using one of two basic evaluation models: push/enter or eval/apply. Implementors use their intuition and qualitative judgements to choose one model or the other. Our goal in this paper is to provide, for the first time, a more substantial basis for this choice, based on our qualitative and quantitative experience of implementing both models in a state-of-the-art compiler for Haskell. Our conclusion is simple, and contradicts our initial intuition: compiled implementations should use eval/apply.