The SMILE model is a Danish, dynamic, data-driven microsimulation model. The current version forecasts demography, education level, socioeconomic characteristics and housing demand for the period 2010-2050. The basic idea with SMILE is to unite the pre-models that the Danish institution DREAM already uses in a full dynamic microsimulation model. The new elements of the model are described and the development strategy is outlined. The model is based on a new Event Pump architecture. This is a Lego-block-like object oriented technique where the model is built as an Agent Tree consisting of Agent objects. The model take extensive use of a method called CTREE, which is a decision tree technique that has not previously been used for microsimulation modelling. Finally, a matching algorithm called SBAM (Sparse Biproportionate Adjustment Matching) has been developed.