The purpose of this paper is:
• Demonstrate the risk of Regression Models generally used in Real State Appraisals, even when following the NB-502/89 recomendations;
• Proposes the usage of more suitable methodology taking in account the influence of important interactions to build Special Models.