Farm-level analysis of trade-offs between soil fertility management alternatives is required to improve understanding of complex biophysical and socio-economic factors influencing decision making in smallholder farming systems and to identify opportunities for improving resource use efficiency. A farm characterization tool (IMPACT) linked to a generic optimization model (Household) was used to evaluate resource use on farms in contrasting wealth categories. The Household model optimized the net cash income for the farms (accounting for all on-farm and off-farm income, costs of production and expenditure for the households). Alternatives for management of nutrient resource were simulated using other models; APSIM for the crop production and RUMINANT for the livestock component. The output from the simulation models was fed into the Household model and evaluated within the biophysical and socioeconomic boundaries of the farms. Analysis of the performance of a poor farmer by IMPACT indicated a yearly net cash balance of US$ -7 per annum (after all needs had been taken care of), mainly due to negative returns from the cropping system. The farmer relied on donated food and fertilizers. The cash balance was negative, even though she also worked for other farmers (i.e. sold labour, about 10 days a month during six months of the crop growing season) to generate income. The net income of the poor farm would be increased to US$81 per annum and the N balance from 7 kg ha-1 yr-1 to 10 kg ha-1 yr-1 by expanding the area allocated to groundnut from the current 5% to 31%. This would, however, generate a huge demand in labour in the current year (extra 46-man days) and reduce the P balance from 0 to -1 kg ha-1 yr-1. Maize could be managed more efficiently on the poor farm by cultivating a smaller, well-managed area. A wealthy farm household with a maize dominated cropping system had a net cash balance of US$210 per annum, mainly from sale of crop products. Under current resource management, the net cash balance could be increased to US$290 per annum by optimization of household energy and protein consumption. The net cash balance for the wealthy farm would be further increased to US$448 per annum, and nutrient balances to 271 kg N ha-1 and 30 kg P ha-1 by expanding the management strategy where maize was grown with a combination of cattle manure and ammonium nitrate fertilizer. To do this, the farmer would need to source more manure (or improve capture and the efficiency with which nutrients are cycled through manure) and invest in 110 man-days extra labour. Expansion of the area grown to groundnut without fertilizer inputs to a third of the farm would reduce net cash balance by US$11 compared with the current crop allocation due to poor groundnut yield. This would also increase labour demand by 155 mandays. Groundnut intensification on the wealthy farm would be more economic and labour effective if a small area was grown with basal fertilizer (7%N, 6%P, 8%K). Despite reducing nutrient balances for the arable plots, feeding groundnut residues to lactating cows increased net cash balance by 12-18% for the current year through increased milk production. The integrated modelling approach was useful for linking biophysical and socio-economic factors influencing decision making on smallholder farms and evaluating trade-offs for resource use in terms of nutrient balances, labour use, food sufficiency and cash balance.