We explored the forecast skill of primary meteorological variables at various steps in the modelling chain for seasonal maize yield anomalies in East Africa and found both potential and real skill for rainfall and temperature in typical cropping seasons. However, forecast skill is a function of geographical region, season, climate variable and forecast lead-time before planting. Next we analyse correlations between historical yields and anomalous weather and climate indicators relevant for arable farming during consequent maize growth stages in two case study regions and found significant levels of correlation and skill that open up the potential for statistical forecasting by use of climate forecast derived indicators. Next, we used a full process based crop model and seasonal climate forecasts to forecast anomalous water-limited maize yield. We find again potential predictability of yields with at least two months lead before planting, in most agricultural regions. On issues of aggregation of simulated maize yields, results showed that enough skill exists, both at national boundaries and at high spatial resolution, to inform maize production related policy decisions at regional or national levels, but also to support maize farming decisions at specific cropping locations.