Crossing and developing inbred lines have been promising options for guava breeding programs. The purpose of this study was to evaluate the genetic divergence among genotypes of S1 inbred guava families by means of the Gower’s technique and the Ward-MLM methodology, to verify the correlation and relative contribution of traits, as well as to identify descriptors with minimum efficiency for this species. The experiment was implemented at the Estação Experimental da Ilha Barra do Pomba, in the municipality of Itaocara, RJ, Brazil. A randomized block design with 18 inbred families, three replicates, and ten plants per plot was used for the experimental design. After 19 months from the planting of the experiment, the 61 earliest and most productive genotypes (individual plants) were evaluated. For this purpose, 29 descriptors were evaluated, of which fifteen were qualitative and fourteen, quantitative. The characteristics required to obtain the distance matrix were analyzed based on the Gower algorithm, and a comparative cluster between the dendrograms of the morphoagronomic variables was achieved from this matrix. Lastly, the Ward-MLM procedure was applied to form the clusters of inbred families. By using all 29 descriptors, greater efficiency was achieved in cluster discrimination. Hence, according to the results identified, it is not possible to indicate minimum descriptors for the culture. Using the Ward-MLM method, the descriptors that most contributed to the divergence among the genotypes were fruit flesh mass, fruit weight, fruit diameter, fruit flesh thickness, fruit placental mass, and fruit length. The most divergent genotypes can be recommended for further crosses or self-pollinations to develop new lines in the guava breeding program of UENF.