NOSQL database management systems adopt semi-structured data models, such as JSON, to easily accommodate schema evolution and overcome the overhead generated from transforming internal structures to tabular data (i.e., impedance mismatch). There exist multiple, and equivalent, ways to physically represent semi-structured data, but there is a lack of evidence about the potential impact on space and query performance. In this paper, we embark on the task of quantifying that, precisely for document stores. We empirically compare multiple ways of representing semi-structured data, which allows us to derive a set of guidelines for efficient physical database design considering both JSON and relational options in the same palette.