Developing countries that employ inflation-targeting monetary policy regimes require accurate short-run food inflation forecasts. We develop a systematic approach to improve food inflation forecasts, apply it to Colombian monthly data from December 1989 to April 2006, and show its relevance for inflation targeting regimes. Forecast accuracy can be improved by: disaggregating food products into away-from-home, processed and fresh foods; employing econometric methods to combine forecasts from individual models; and using flexible least squares methods in the presence of structural changes. We also show the importance of accurate food inflation forecasts in models simulating monetary policy transmission.AMS Subject Classification ,, Copyright 2012, Oxford UniversiTY Press.