Precision agriculture in cropping areas of dryland Australia has focused on managing within production zones. These are ideally stable, possibly soil- and topography-based areas within fields. There are many different ideas on how to delimit and implement zones, and a four year whole-field experiment, with low, medium and high treatment philosophies applied per 9m seeder/harvester width across the entire field, was established to explore how zones might best be established and used. The treatment philosophies combined wheat/barley seed rate, starter (N/P) and in-season fertilizer, varying according to season, but applied to the same seeder widths over the four years. The boundaries of treatment responses, determined using geographically weighted regression on yield data in each season, were compared to apparent electrical conductivity (ECa) and elevation surveys, and also zones created from clustering prior yield data. There were significant responses to the medium or high treatments over 23-66% of the field area in three out of four seasons. The zones created from yield data were good at predicting the pattern of yield variation, but not treatment responses. Responses were most likely to the fertilizer component of the treatments, and related to field history, and nutrient use in previous seasons. The pattern of ECa and elevation was similar to the yield zones, and not well related to treatment responses. The results support the role of fixed zones related to yield as a way of estimating overall input requirements and limits of likely responses, but imply that in-season crop sensing will be required to predict the responsive parts of fields, which will change from season to season.