User models are important tools for personalization, especially in ecommerce applications. But capturing dynamically changing user needs is a challenge. One of the reasons for such difficulty is that the purchasing behavior of an individual is based on a number of different aspects. In this paper we identify these aspects as a combination of demographics, domain based expectations and transactions. Since each individual can demonstrate a unique combination of these aspects, to achieve finer personalization, such individuality will have to be captured in user models. In this paper we propose such a user model architecture, which also has the ability to self-improve adapting to changes of individual behavior and long term modeling possibilities.