This paper focuses on a topic which was not supposed to take such importance within the DÉRom: data structuring in the different entries. The structure of the lexical material is not a mere representation of the phenomenology of the Romance lexicon; it is supposed to follow the historical evolution of the discussed lexeme. This is done by reconstructing of (formally and historically unitarian) sub-etyma which may in a second move, by internal reconstruction, lead to a more general etymon. Although we first supposed subdivided entries would be exceptional, a significant part of the DÉRom entries are now organised by sections: this organization became the main way of understanding and writing (sociologically and geolinguistically) the history of the examined lexemes, since the articulation of the entry reflects the more important events in the history of the word. Although data structure designing is often the last step in the lexicographical work, it can lead to completely reexamining the data (for instance, when a category is almost empty or contains geolinguistically incoherent units, this can be an argument for a different analysis - as a borrowing for example). We show, through a set of examples, the many ways data structuring interacts with etymological interpretation and expose some of the difficulties in designing such a structure. We adress the following central questions: (1) How can comparative linguistics be used as a starting point for a multilingual dictionary project? DÉRom is a multingual dictionary project whose aim is comparative linguistics, the purpose of each entry being reconstructing a Proto-Romance lexeme. Linguistic comparison is the way for organizing the structure of each entry, and this is the point we will adress specifically. (2) Which ways are there to a comparative historic lexicography? We explain the specific way of presenting the material we had to develop during the compiling of our historical comparative dictionary.