Your starting point
Prerequisites: Use the ideas and vocabulary from lesson 3; revisit its worked examples if needed.
Definition: Provenance is the recorded origin and history of data: who collected it, how, when, for what purpose, and what transformations followed. A primary source records the observation directly; a secondary source reuses or summarizes it.
Follow the chain of evidence
A number copied into a presentation can lose its original definition. Trace it back to a source document or dataset, not merely another presentation that repeats it. Record the collection period, population, inclusion rules and version. A fresh download does not guarantee fresh observations. The publication date and the measurement dates may be different.
Ask why the data exists
Administrative records are often collected to run a service, not to answer your research question. A support system records people who contacted support, excluding customers who had a problem but did not report it. An optional survey includes people who chose to respond. Both can be useful, but neither automatically represents every customer. Relevance depends on the decision and population you care about.
Preserve an audit trail
Keep an unchanged source copy and describe each transformation. A renamed column, removed duplicate or imputed value is part of the analytic history. Document assumptions that cannot be verified. AI can suggest questions about provenance, but an invented source URL is not documentation. If a source cannot be checked, narrow the claim or postpone the conclusion rather than laundering uncertainty through a polished summary.
Two worked examples
A survey headline
A fictional newsletter says 90% prefer online workshops. The underlying survey sampled 90 of 100 respondents from a mailing list for online events. The fraction is calculated correctly, but it describes respondents from that list. It cannot establish the preference of all local residents without a defensible sampling argument.
Two versions of sales
An April export records revenue before returns; a later export subtracts refunded orders. The files can differ even with identical date filters. Preserve both versions and their extraction times. A comparison that labels the difference as sales growth would confuse a definition change with a change in customer behavior.
Where this becomes useful
A nonprofit wants to use public transport counts to plan outreach. The source may count boardings, so a person changing buses can appear twice. Provenance determines whether those counts fit a question about unique people reached.
Make an attempt
Open the synthetic Public transport table. Write a source note that includes the fictional status, row unit, units and known limitations. Identify one detail you would need before making a real transport policy recommendation.
Misconception check
Claim: An official-looking chart is enough evidence.
Inspect the correction
Inspect the source and definitions behind it. Attractive formatting does not establish collection quality or relevance.
Independent application
Choose a public, non-sensitive dataset description without downloading personal records. Write a provenance card with owner, measurement dates, population, units, version and one limitation. State one question it could answer and one it cannot.
Completion records practice and self-review. It is not certification of statistical competence.
