The U.S. Bureau of Labor Statistics (BLS) is using DDI to document Consumer Expenditure Surveys (CE). CE produces PUMD files on a yearly basis. The immediate goal in the development of the system is to allow users to connect PUMD variables to their questions. The current stand-alone database with the information on all the variables is incomplete, and the ability to track changes to codes, definitions, variables, and files over time and across surveys is not straightforward. Our plans are to incorporate functions to view the mapping of variables and questions and track changes. In addition, we want to record methodological documentation such as questionnaires, weighting and sampling designs. In the iteration of the metadata system currently under development, we are mapping the variables from the 2017 PUMD files to their questions, which either wholly or in part provide the input to them. In this talk, we describe the development effort with particular emphasis on variable and question mapping. In many discussions about metadata models, the mapping between variables and questions is glossed over—the implication being that mapping is straightforward. Our experience is different, and we describe the difficulties faced.