Data from a long-term trial on rates of fertilizer nitrogen (N) application to maize (Zea mays) were used to validate a cropping systems simulation model (APSIM) and then to apply the model to explore the risk associated with N fertilizer use by smallholder farmers and management strategies to minimize that risk. On average, maize growth and development in response to N was simulated with a degree of accuracy that justified its use in analysis of risk associated with N use in these semi-arid regions of Zimbabwe. APSIM was then configured to simulate the response to N over a 46-year climate record in order to assess the long-term risks associated with N use. The simulated long-term distribution indicated that negative responses to N could be expected in 15% of years, whereas no negative response to N was recorded in the experiments at the Makoholi Research Centre. Median responses were 20–30 kg maize grain kg−1 N for observed and simulated results. In terms of return on fertilizer investment, the observed and simulated distributions were also similar: in about 20% of years, a negative return could be expected, while in the best 20% of years a return of $Z5000 or more could be expected given the grain:fertilizer price ratio which, at March 2000, was about 1:7. The model analysis has suggested moderate rates (approximately 30 kg N ha−1) of N fertilizer would give greater responses per unit N applied than smaller rates (15 kg N ha−1). There was no evidence that conditional fertilizer strategies based on early-season rainfall would offer significant benefits over fixed application strategies. Early sowing at recommended population densities gave higher responses to N than were achieved for late sown or low-density crops.