Some of today’s most widely spread applications are social systems where people can form communities and share knowledge. However, knowledge sharing is not always effective and communities often do not sustain. Can user modelling approaches help to identify what support could be offered and how this would benefit the community? The paper presents algorithms for extracting a model of a closely-knit virtual community following processes identified as important for effective communities. The algorithms are applied to get an inside of a real virtual community and to identify what support may be needed to help the community function better as an entity.