Cluster analysis has been widely used to delineate soil and crop management zones. Since results of noninferential cluster analysis are totally dependent upon the variables used for similarity-dissimilarity measure, the addition or deletion of relevant variables can substantially impact the resulting solution. We evaluated clustering approaches using different data for management zone delineation to determine which approach best captured spatial variability of soil nutrient levels and crop yields. Research was conducted in four fields in two physiographic regions of North Carolina. Nonhierarchical k-means cluster analysis was performed to delineate zones based on commonly used clustering variables: apparent soil electrical conductivity (EC_a), bare-soil near-infrared (NIR) radiance, elevation, slope, and their combinations. The strengths and signs of correlations between clustering variables and soil test P, K, pH, CEC, and humic matter (HM) varied across fields. Relationships between clustering variables and crop yields were inconsistent among years and fields. Most of the clustering approaches effectively minimized the within-zone variability of soil test and crop yield, but the degree of variance reduction varied considerably depending on fields as well as clustering variables used for zone delineation. Of the soil-based management zone strategies, cluster analysis on the combination of EC_a and NIR consistently captured more variability of K, CEC, and HM in the coastal plain fields, while the efficiency of zone approaches varied markedly with clustering variables in the piedmont fields. Differences between zone delineation approaches to capture yield variability were minimal. Efficiency in capturing yield variation depended more on temporal variability of crop yields than on zone approaches. Our results suggest that zones created by cluster analysis could provide a way to group and manage spatial variability of soil nutrients within fields. However, appropriate selection of clustering variables for an individual field could be critical to maximize effectiveness of management zones for site-specific management.