Diversity of farming activities may increase income stability and reduce risks to resource-poor households, while integration-using the outputs of one activity as input in another activity-may reduce dependency on external resources. In practice, diversity and integration are poorly defined, and there is no method to characterise them, hampering the exploration of their benefits. We introduce a method based on network analysis (NA) to assess the diversity and integration in farm household systems by using the Finn cycling index to characterise integration of farming activities, and the average mutual information, and the statistical uncertainty (H _R) to characterise diversity of flows. We used nitrogen (N) flows in an application of NA to crop-livestock systems of the highlands of Northern Ethiopia. N recycling was low (FCI <3%) in these systems independently of the farm type. Even with improved N management FCI was lower than 10%. Since large amounts of N are withdrawn from the system with the harvests, there are relatively few opportunities for recycling. The diversity in N flows increased from the poor to the wealthier farm households, but differences were small. The wealthier farm households did not recycle more N than the relatively less diverse and poorer farm household. The definition of the system and compartments boundaries must be made explicit in any application of the method as these strongly affect the results. NA appears useful to assess the effects of farm management practices on system performance and to support discussions on diversity and integration of agro-ecosystems.