Information systems have evolved into complex data platforms supporting end-to-end data-intensive needs, aimed at coping with the different V's that characterize Big Data. In particular, multi-model databases (MMDBs) have been proposed to natively support storing and querying data in different (schemaless) models, so as to better handle Variety. In this work we envision a new data warehouse architecture in which an MMDB is used to enable on-the-fly user-driven extensions of multidimensional cubes with additional data, while ensuring support to variable and complex data and keeping the impact on ETL low. After proposing the architecture with the aid of a case study on the management of emerging plant disease, we discuss the main associated open issues.