Quality Assurance
Error reports: the ones you receive are a fraction of the ones that exist
Most people who notice an error do not report it. So each report received should be treated as a sample, not an incident.
#quality assurance
Quality Assurance
Quality Assurance
Auditing a pipeline: sample by hand, because automation checks format not truth
Automated checks confirm that data is well-formed. They cannot confirm that it is right.
Quality Assurance
Uncertainty: say what you do not know rather than presenting a clean number
Academic data is uncertain in ways that do not show in the output. Presenting it cleanly is a choice, and usually the wrong one.
Quality Assurance
Institutional records: the institution is the authority on itself
Institutional data drifts constantly through renames, mergers and restructuring — and only the institution knows what actually happened.
Quality Assurance
Trustworthy data: provenance on every field, or it is just a list
What separates a registry from a scraped list is not accuracy. It is being able to show where every value came from.