Document stores are frequently used as representation format in many applications. It is often necessary to transform a set of data stored in a relational database (RDB) into a document store. There are several approaches that execute such translation. However, it is difficult to evaluate which target document structure is the most appropriate. In this article, we present a set of query-based metrics for evaluating and comparing documents schemas against a set of existing queries, that represent the application access pattern. We represent the target document schema and the queries as DAGs (Directed Acyclic Graphs), which are used to calculate the metrics. The metrics allow to evaluate if a given target document schema is adequate to answer the queries. We performed a set of experiments to calculate the metrics over a set of documents produced by existing transformation solutions. The metric results are related with smaller coding effort, showing that the metrics are effective to guide the choice of a target NoSQL document structure.