Soil water retention capacity is usually estimated by pedotransfer functions which, in turn, are very problematic to apply for soil maps containing only categorical-type soil properties. Most of the soil information in Hungary is available in the format of soil maps reporting categorical-type information and often lacking data about measured hydrophysical properties. Here we are mainly concerned with the possibility of providing predictions of soil water retention categories starting from continuous- and/or categorical- type information. First we used Factor Analysis (FA) to synthesize the information contained in the soil variables. We then employed the factors extracted in the context of cluster analysis (CA) to differentiate soil groups that can be characterized by means of typical soil water retention values. Finally, we used multinomial logistic regression to address the relationship between the soil water content and the other soil properties used in FA.