Within the Big Data trend, there is an increasing interest in Not-only-SQL systems (NoSQL). These systems are promising candidates for implementing data warehouses particularly due to the data structuration/storage possibilities they offer. In this paper, we investigate data warehouse instantiation using a document-oriented system (a special class of NoSQL systems). On the one hand, we analyze several issues including modeling, querying, loading data and OLAP cuboids. We compare document-oriented models (with and without normalization) to analogous relational database models. On the other hand, we suggest improvements in order to benefit from document-oriented features. We focus particularly on extended versions of OLAP cuboids that exploit nesting and arrays. They are shown to work better on workloads with drill-down queries. Research in this direction is new. As existing work focuses on feasibility issues, document-specific implementation features, modeling and cross-model comparison.