In the context of big data where the volume of data is continuously growing, we propose in this paper an agile approach to schema evolution in data warehouses that allows designers to integrate new data sources and take into account new user needs in order to enrich the analysis possibilities. Our approach is based on a multi-versioned evolutionary model using graphbased assembly modelling. A meta-model will allow the management of the versions of the warehouse schema. We also propose evolution functions at the schema level (the evolution of instances is not dealt with here). We validate our approach with a software prototype that manages both versions and schema evolution in the model.