# Export the evidence, not just the attractive column

A useful research CSV should survive being separated from the dashboard that explained it.

By AppRill Editorial · Published 2026-09-08
Canonical: https://apprill.app/blog/research-export-provenance

You export a list of apps and revenue figures. A week later, nobody remembers which country was selected or whether the numbers were estimates.

The export succeeded technically and failed as a research artifact.

## Keep the context with the value

Include stable app identifiers, platform, source, metric name, unit, geography, measurement period, observation time and evidence type. A monetary value should retain currency and the distinction between observed, reported, estimated and calculated.

For a hypothetical comparison, a blank revenue field should remain unknown. Converting it to zero can reverse the meaning of the record once somebody sorts the spreadsheet.

## Make the file understandable alone

Use descriptive column names and a short data dictionary. Preserve links to source records where sharing is permitted. If a field has important limitations, include a companion note rather than relying on a tooltip that exists only in the web app.

Be careful with spreadsheet interpretation of untrusted text. Exporters should handle values that spreadsheet software might treat as formulas, and preserve identifiers that contain leading zeros or exceed ordinary numeric precision.

## Respect the sharing boundary

An export can include private collections or account-specific notes. Confirm which fields belong in the file and who can access it. A public research summary should not accidentally carry customer identifiers or internal operational details.

For licensed data, check permitted export and redistribution terms before publishing a downloadable dataset. The presence of an export button does not automatically grant every downstream use.

Finally, reopen the file in the intended tool and compare a few records with the source. Verify that dates, currencies, blanks and identifiers survived.

A good export carries enough context that another person can interpret the evidence without borrowing your memory. That is the difference between a useful research handoff and a spreadsheet full of confident-looking mysteries.


## Sources