Robot Data Revisions: How to Tell a Market Change From a Method Change

A robot statistics table can change between reports without the underlying factories changing on the same day. Suppliers may revise submissions, a classification may be clarified, or the publisher may update its method. When a new value replaces an earlier one, a data user should first determine whether the difference is a real observation or a change in the record behind it.

Keep the original publication with the revision

Record the report title, publication date, reference year, metric and version of the table. If a later release changes a historical value, keep the earlier value in a historical field and store the new value with its own source. Do not overwrite the first record without a note, because a reader may need to reproduce an earlier analysis made from the original publication.

The International Federation of Robotics' industrial robot release presents annual installations and operational stock as different measures. A revision to one measure should not silently alter the other. The source's year, unit and definition belong beside the number in the data sheet.

Ask whether the denominator changed

A percentage or density can move because the numerator changed, the denominator changed, or both changed. A regional share of new installations is different from a density calculated per manufacturing employee. If a table changes the industry scope, employment series or geographical coverage, the new percentage may not be comparable with the old one even when the label looks similar.

Store the numerator and denominator where the source provides them, along with the population or industry definition. If the source publishes only a rounded share, label it as rounded. Avoid reverse-calculating a precise number from a rounded percentage and a separate total unless the assumptions are stated.

Separate a correction from a new forecast

A historical correction and a new forecast are different events. A corrected 2024 installation value changes the historical record; a 2025 projection expresses an expectation made at a particular time. When a chart contains both, mark the status and publication date so readers can see what has happened and what remained uncertain at the time of publication.

The IFR 2025 presentation marks future installation values as forecasts. A later observed total should be compared with the forecast vintage, not substituted into the original chart without preserving its status. A forecast error is informative only when the two figures use matching definitions.

Track the reason for a changed value

If a publisher provides a methodology note or revision explanation, link it to the affected rows. Write a short reason such as a supplier correction, a scope change or a reclassification, and identify whether the reason is confirmed by the source or inferred by the analyst. An unexplained difference should remain unexplained in the published data rather than being assigned a likely cause.

Keep a copy of the relevant footnote with the row rather than assuming a reader will find it from the report title alone. Footnotes often describe exclusions, rounding or a change in sample and can materially alter the meaning of a percentage. If the note does not apply to every row in a table, record the affected rows explicitly.

Where the source gives no revision history, compare the table structure and footnotes and contact the publisher if the decision depends on the difference. Keep the response and date of inquiry with the research file. A lack of explanation is itself a limitation that should be visible to a reviewer.

Make downstream calculations reproducible

For each derived chart or growth rate, store the input rows and formula. Note whether the inputs are observed, forecast, revised or rounded. If a historical series is recalculated after a revision, update the chart title or note so readers know which vintage it uses. Do not compare a revised historical value with an unrevised value from another release and call the difference a market trend.

When a value is copied between a spreadsheet and an article, carry its source and vintage into the editorial notes rather than leaving them in a separate folder. This small practice makes a later correction easier: the editor can locate every statement that used the old value and decide whether the conclusion still holds.

Use a change log that records the old value, new value, source, date and reviewer. This is especially important when multiple analysts or sites reuse a central dataset. A clean change log can prevent an accidental copy of an obsolete figure into a new article.

Report the boundary with the conclusion

A defensible data story says which release supplied the figure, what it measures, and whether it was revised or projected. It avoids implying that every historical discrepancy is a market event. For related measurement boundaries, read our forecast-versus-observation guide and browse the robot statistics archive.