The approach usually taken when conceptualizing metadata is typically one of 'documentation', which supposes a reference object, e.g., data, and something to be told about it. That documentation is static; it describes data in a specific state, usually as ready for publication. The life-cycle idea introduces a new perspective. Sure, one can still reduce metadata to a report about what happened across the life cycle of the data. But there is also an opportunity to model metadata in a way to support work done to obtain data, process, edit and publish them. The following example will be developed. Because we don't use the integrated tools for handling metadata and data all over the life cycle, consistency between the two levels of information may be broken. We need a metadata model, which supports the comparison of metadata drawn from more than one source, e.g., the questionnaire, treated as 'metadata by design', and information extracted from a semi-documented data file, an SPSS data file for example, which stands as the 'fielded metadata'.