Data provenance – the foundation of data quality

Provenance for data is often defined by analogy with its use for the history of non-digital artifacts, typically works of art. While this provides a starting point for our understanding of the term, it is not adequate for at least two important reasons. First, an art historian seeing a copy of some artifacts will regard its provenance as very different from that of the original. By contrast, when we make use of a digital artifact, we often use a copy, and and copying in no sense destroys the provenance of that artifact. Second, digital artefacts are seldom “raw” data. They are created by a process of transformation or computation on other data sets; and we regard this process as a part of the provenance. For “non-digital” artefacts, provenance is usually traced back to the point of creation, but no further.