Three pages are not always three independent sources

If a company announcement is copied into a directory and summarized in a newsletter, you have three pages—but possibly only one origin for the claim. Counting links alone can make weak evidence look well supported.

This fictional example shows how to record that distinction. It is a research worksheet, not a claim about any real company.

Worked example: does a product support CSV export?

• Page A: a vendor announcement says CSV export is available.
• Page B: a directory repeats that announcement.
• Page C: a newsletter cites the same announcement.
• Page D: a separate reviewer documents their own export test.

Four pages, two recorded evidence origins: the announcement and the separate test. Only the test is an independent observation in this example. Neither proves the feature is still available today or included in every plan.

Download and adapt the worksheet

Download the fictional CSV. Each row records a page, its evidence origin, the kind of evidence, and whether it independently tests the claim. The origin IDs are manually assigned after checking citations; they are not a truth score.

A practical review sequence

1. Write the exact claim, including product version, plan and date when relevant.
2. Follow citations back to the earliest accessible evidence. Do not assume different domains mean different origins.
3. Distinguish repetition, a vendor statement, and a separate observation.
4. If the origin is unclear, record unknown. Do not merge all unknown origins into one source or count each as independent.
5. Look for contrary evidence and check whether the result applies to your intended use.

An official source can be the best authority for a policy or price. Independence is useful for corroboration, not a replacement for relevance, authority or freshness.

Get the worksheet and executable example on GitHub, or use the dataset-buyer checklist before purchasing data.