What to record when a public dataset quietly changes column names

When a public dataset changes column names without a loud announcement, record the old name, new name, access date, and any effect on reused summaries.

Column names look like technical details, but they can change the meaning of a chart or written claim. A field called “registered_users” may become “active_accounts,” or “revenue” may split into gross and net columns. A reused notebook, spreadsheet, or article can keep running while silently reading a different concept.

A useful source note records the dataset title, publisher, file or table name, access date, version or release number if available, old column name, new column name, and whether the values changed or only the label changed. If the publisher provides a changelog, link it. If no changelog exists, say that the change was observed during access.

The downstream check matters. Identify every chart, quote, model feature, or summary sentence that used the old column. Some will only need a label update. Others may need recalculation because the column definition changed.

The practical rule: do not hide schema changes in cleanup notes. If a dataset column affects a public conclusion, the change belongs in the source trail so future readers can understand why an older chart or number looks different.

What to record when a public dataset quietly changes column names

// COMMENTS

ON THIS PAGE