During the last few years, endless data have supplied the marketing area with new perspectives for the definition and segmentation of markets, in addition to the company's private and internal data banks as well as of public and external ones. An example of the latter are the virtual communities, and Orkut is one of its exponents. This article, offers an exploratory approach to market segmentation using two alternative methods (logistic regression and neural networks) through secondary data collected from Orkut, assuming the possibility to predict certain consumption attitudes described by members of virtual communities. This way, two communities self denominated I love cold beer and I hate beer were picked up, and each one supplied random probabilistic samples among 400 members. Processing and analyzing the data collected in three phases - data cleaning and selection of variables of interest, discriminating analysis, and analysis through neural networks - confirmed the possibility of determining consuming attitudes, and consequently of using the Orkut virtual community as a data bank for segmentation. Further contributions, limitations and implications are part of this study.
segmentation; marketing; virtual community; neural networks; logistic regression
