Precision agriculture has provided farmers to obtain and deal with variations which have been found in an agriculture area. The management zones (MZs) help to set precision agriculture up in a more feasible and economical way. Thus, this trial aimed at generating MZs with different numbers of variables by fuzzy C-means algorithm, based on attributes copper, silt, clay and altitude, which were correlated with soybean yield. Its main goals are: to evaluate the relative efficiency of each MZ and determine if the use of more than one variable correlated with yield could provide the generation of more effective and economical MZs. Fifteen MZs designs were generated and divided from two to five classes. The variable with the highest spatial correlation with yield and at the same time had spatial self-correlation could be considered appropriate on MZ generation process. Therefore, there was no need to use other redundant variables. The experimental design based only on copper variable, divided into two classes, was the most suitable for yield management.