Data Quality Assessment (CPP-019)

Definition and scope

Data Quality Assessment refers to the systematic evaluation of Objects and their associated Metadata against predefined measures to ensure they meet the standards necessary for consumers' needs and continued access. The assessment typically covers several key dimensions, some of these are for example:

• Authenticity: The Object is what it purports to be (i.e. it has been created, modified and sent by the person purported to have done it at the date and time purported). The designated community must be able to trust that the data is real and credible and is managed by a trustworthy TDA. Sufficient information must exist to understand the Object's creation circumstances, provenance, and relationship to other content. In addition to integrity checks, the authenticity of the data is ensured by controlled changes through preservation actions and the Provenance metadata.
• Completeness: The Object and the Metadata are complete. They do not have missing parts or links to targets outside the preserved Object which should remain accessible.
• Consistency: The Object is presented in applicable file formats or Representations with applicable metadata formats. Conflicting values in the Metadata should be avoided.
• Relevance: The data preservation is based on a predefined collection development policy (i.e. has a purpose of being preserved).
• Structured: The structure of the Object is described in the Metadata. Complex Objects are organised, including relationships between Files, proper sequencing of multi-part Objects, and the integrity of any embedded Metadata or links.
• Understandability: The information is understandable and meaningful for the designated community.
• Validity: The Object and Metadata are valid against the File and metadata format specifications and standards, and comply with all other predefined profiles and rules.

Data quality assessment may include various processes, repeated from time to time. A very common phase to perform an assessment is in the Ingest phase, but the use case described below demonstrates that such a process can also be performed at the access stage. The data may be rejected from digital preservation, if it does not meet the criteria. An assessment typically has the following steps, described on a very high level:

1. Define the scope of the assessment;
2. Define data quality dimensions and metrics, including possible thresholds;
3. Gather and analyse data;
4. Create a quality report about all the findings;
5. If needed, update the data and Metadata to improve the quality.

The step-by-step description below mainly concentrates on the technical aspects of the data and Metadata, but the scope of this CPP is indeed covering a broader range of contextual data quality properties.

The assessment process often employs both automated tools and manual review. For example, automated tools can perform file format identification, validate File or Information package structures, check for malware, verify checksums, or check for completeness of a delivery against an inventory. Human reviewers, for example, may evaluate content accuracy, Metadata completeness, and contextual adequacy. The processes should be automated as much as possible for faster processing and to avoid human errors.

Results from Data Quality Assessment affect preservation planning decisions (e.g. what additional Metadata needs to be captured). The assessment also establishes baseline quality metrics that can be monitored over time to detect degradation or other changes that might necessitate intervention.