How to cite data sources in a chart

A useful chart source line lets a skeptical reader identify where the numbers came from. “Source: internet” fails. “Source: Bureau of Labor Statistics” is better, but may still be too broad when the agency publishes thousands of series. The strongest record names the producer, dataset or series, relevant table or identifier, and a route back to the underlying data.

Traceability matters for more than academic formality. Economic data gets revised. Company filings are restated. Dashboard URLs change. A colleague may need to update the chart six months later without reverse-engineering your spreadsheet. Good source metadata makes verification, correction, and reuse faster for the entire editorial team.

The visible line on the chart can remain concise. Preserve the full citation and transformation history in the article caption, footnote, accessible description, CMS, or internal chart record. Think of the on-image source as the label and the full record as the audit trail.

Citation styles differ, and many publishers have house rules, but the underlying provenance fields are stable. Capture them once, then format a shorter or longer citation for each publishing surface.

The source identifies the origin of the data. The method explains what your team did to it. A visual credit identifies who designed or published the graphic. Combining all three into one vague line makes the chart harder to audit.

For example, a mortgage-rate chart may use weekly observations from a Federal Reserve database, calculate annual averages, and be designed by your newsroom. The data source did not create your annual values, and the designer did not produce the underlying survey. Preserve those roles: “Source: Freddie Mac via FRED. Method: annual average of weekly observations. Chart: Example Research.”

If you recreate someone else’s chart from their underlying dataset, cite the data producer and consider crediting the original analysis if its selection or transformation was substantive. If you reproduce the original image itself, copyright and permission requirements may apply in addition to citation.

The exact wording should follow your publication’s style, but these patterns are compact enough for charts and specific enough to help readers verify the work. Replace bracketed text with the most precise information available.

Most published charts do more than plot a downloaded column. You may calculate a moving average, index values to 100, adjust for inflation, divide by population, group categories, or combine two providers. Those choices affect the numbers the reader sees and belong in a methodology note.

Keep the source line focused on provenance: “Sources: BLS; Census Bureau.” Then describe the transformation: “Monthly employment divided by annual population estimates; 12-month moving average.” If the calculation is central to the claim, include enough detail for an informed reader to reproduce it or link to a fuller methodology.

For estimates or manually collected data, identify judgment calls. Explain how missing values were handled, how categories were coded, and whether figures were transcribed from documents. If the chart contains both reported and estimated values, distinguish them visually and in text.

Many official series are revised after initial publication. Employment, GDP, trade, and survey estimates may change as more complete information arrives. A chart published today can therefore disagree with the same database queried later even when both were correct at the time.

Record the retrieval date for continuously updated sources and the release or vintage when it matters to the story. If you are analyzing what policymakers or markets knew at a historical moment, use a vintage database where available rather than today’s revised history. If a revision materially changes a published conclusion, update the chart and add a correction or revision note according to your editorial policy.

URLs alone are not enough. Query parameters, API series IDs, spreadsheet table numbers, and downloaded filenames often contain the detail needed to reproduce the extract. Keep them in the internal record even when the public source line remains short.

A chart embedded in an article can rely on a caption and linked methodology. A social graphic may circulate as a screenshot with no surrounding copy, so its source must travel on the image. A slide deck can use a short footer plus a references slide. A formal report may need numbered figure notes and a bibliography entry.

Do not shrink the source until it is technically present but functionally unreadable. If space is tight, shorten organization names consistently, remove protocol text from a displayed URL, or use a brief source on-image with a fuller link in the post. The source should remain legible at the final exported size.

Giving credit does not automatically grant permission to reuse a dataset or an existing chart. Public-domain government data, open-licensed datasets, copyrighted databases, and commercial market data have different rules. Read the provider’s terms, including requirements for attribution, share-alike licensing, non-commercial use, or prior approval.

FRED, for example, distributes series from many original providers and tells users to inspect the notes and tags for suggested citation and copyright status. The same platform can therefore contain public-domain data and copyrighted series with additional restrictions. When in doubt—especially for commercial publishing—confirm the rights with the provider or your legal/editorial lead.

If you use a screenshot or reproduce someone else’s finished visualization, you are using both the underlying data and a designed work. Rebuilding the chart from properly licensed data may simplify the rights question, but you should still credit distinctive original analysis when appropriate.

Source review should happen before export, not after a visual has been approved. Make one person responsible for checking provenance and method against the final data. This is particularly important in semi-automated publishing workflows, where a template can make an unverified number look publication-ready in seconds.