The internet has become a big battlefield where organizations are trying to keep their present clients and to gain new ones. Two important weapons that the organizations have are to make a good web site design and to have a content interesting for the visitors. To improve the web site content many tools have been developed. However, it is hard to figure out how to apply these changes. Furthermore, in complex web sites, this is a non trivial task. We propose a novel approach that helps to improve a web site content using a SOFM and performing a reverse clustering analysis that allows us to gather the most representative web pages from a web site, using this small set of pages as a guideline of how these enhancements should be performed. The effectiveness of the method was tested in a real web site.