Colquhoun.AI
← All guides

Research Guides

Match Dataset Downloads to the Release the Page Describes

· Colquhoun.shop editorial

A dataset landing page says “download the research,” but its CSV and spreadsheet links contain different releases. Readers can reach both files and still be unable to reproduce the article's figures. The missing check is not link availability; it is agreement between the description and each downloadable distribution.

This release worksheet is for publishers preparing an actual dataset they are authorized to share. It does not announce a Colquhoun.shop dataset or describe a dataset-hosting feature.

1. Distinguish the dataset from its files

Google's Dataset documentation distinguishes the landing-page URL from a distribution's download location. It describes distribution.contentUrl as the download link and distribution.encodingFormat as the file format. Multiple distributions can represent different downloadable formats. Its metadata guidance supports dataset discovery, not verification of the underlying measurements. Google: Dataset structured data

Use that distinction in the editorial review. Decide what collection the release represents before listing its files. A spreadsheet containing extra analysis is not necessarily an interchangeable copy of a plain data export.

2. Write a small release contract

Record the release identifier, included observation period, row meaning, field definitions and known exclusions. This is an internal acceptance record, not a legal contract. Use it to decide whether each download belongs to the release described on the page.

For every format, state whether it contains the same records, a documented subset or additional material. If the workbook includes charts, formulas or a notes sheet, explain which sheet contains the underlying data and which parts are derived.

3. Catch a plausible mismatch

Imagine a fictional website study with 48 observations in its November release. The landing page describes those 48 observations. The CSV contains the current release, but the workbook link still serves October's 42-observation file.

Both links load successfully. Both filenames look reasonable. A reviewer catches the mismatch by checking the release identifier and observation range inside each artifact, then comparing the expected record set. Counting rows alone would miss a different error where both files had 48 rows but represented different observations.

The publisher replaces only the incorrect distribution after review and reruns the checks. The earlier artifact remains identifiable in the internal release record, so the correction does not erase the history of what was served.

4. Build a distribution acceptance matrix

Keep the visible download labels, release notes and machine-readable description aligned. A metadata validator cannot tell you that a workbook silently contains an older study. The artifact review must happen independently.

Check the proposed sharing terms with the responsible owner and describe only the permissions actually granted. Do not assume that a public download allows every form of reuse, or expose private source records merely to make a dataset look complete.

5. Rehearse the reader's route

Ask a colleague who did not prepare the release to start at the landing page, choose a format and identify the observations supporting one published figure. Give them the public instructions, not private explanations that ordinary readers will never see.

Record where they hesitate. If they cannot distinguish the raw data from an analysis sheet, improve the labeling. If the article cites an earlier release, provide enough version context to explain that relationship instead of implying every download reproduces every historical claim.

A release is ready when its description, distributions and evidence agree within documented limits. This makes reuse easier and a citation more useful, but it does not promise backlinks, indexing or independent validation of the research.

This guide describes the built-in Web Reader and Keyword Density behavior verified for this publication. Available data depends on the page response; these tools do not measure visits or search rankings.