This work assesses the value of climate forecasts for millet growers in Niger. We quantify the potential value of three kinds of categorical forecasts assigning cumulated rainfall during the forthcoming rainy season by tercile (dry, normal, or humid): (i) a realistic imperfect tercile forecast; (ii) a perfect tercile forecast; and (iii) a perfect tercile forecast which includes a prediction of rainy season onset and offset dates. Eighteen management strategies are assessed. Corresponding yields are computed using the SARRA-H crop model then converted into utility by taking into account risk-aversion. Simulations over an historical 18-year period 1990-2007 show that benefit is lowest with imperfect tercile forecasts (+6.9%), higher (+11%) with perfect tercile forecasts, and reaches +31% when enhanced adaptation strategies and additional climatic indices are available. These results show that improving existing forecast systems by including the prediction of onset and cessation of rainfall is of great value.