The difficulty of beating the random walk in forecasting spot foreign exchange rates is well documented. In this paper, we propose a functional principal component-based scalar response model which we benchmark versus leading VECM frameworks. Our approach leads to near systematic outperformance in terms of a comparison of performance measures, and to multiple instances of statistically significant improvements in forecast accuracy. Overall, our results provide evidence that the forward rate term structure contains substantial information about the evolution of the spot exchange rate. Finally, a stylised trading strategy is employed to demonstrate the potential economic benefits of our approach.