The ubiquitous access to information via the Web has changed our daily life and business practices. E-commerce applications allow for a wide variety of vendors to compete in a world-wide market. But with a growing number of vendors, also the amount of available offers is increasing, which in turn leads to an information flood that may severely hamper the user experience. Skyline algorithms as introduced by the information systems community, promise to winnow suboptimal offers from electronic marketplaces. For the amount of Web data and the typical interaction style, however, the result sets are still too large and hard to manage. Recently, first approaches to integrate human decision processes like compromises or trade-offs have been designed. In this paper we will build on these approaches and introduce a novel heuristic into the skylining paradigm that not only allows for convenient Web-style user interaction, but also focuses searches on semantic clusters of offers. Thus, the view on interesting clusters is refined, whereas less interesting clusters are strongly reduced in size and strictly focused on only the outstanding items.