Cotton in Brazil is an input-intensive crop. Due to its cultivation in large fields, the spatial variability takes an important role in the management actions. Yield maps are a prime information to guide site-specific practices including delineation of management zones (MZ), but its adoption still faces big challenges. Other information such as historical satellite imagery or soil electrical conductivity might help delineating MZ as well as predicting crop performance. The objective of this work was to evaluate the importance of numerous sources of information in predicting the crop development and delineating MZ in cotton. In order to represent the historical data set, 16 variables were chosen, including satellite imagery collected over the 27 years before the crop season, soil electrical conductivity and digital elevation model. Four variables represented the in season data set: satellite and terrestrial-derived NDVI and sensor-derived crop height. Both data sets were resumed to two latent variables from canonical correlation analysis. Historical and in season variables showed to be highly correlated (r=0.87) indicating that in the lack of in season crop development data, historical information can be used to delineate management zones and guide site-specific management strategies.