Styxis
Company
Knowledge

Distinguish visitor change from a tourism data-source revision

A revised historical series and a change in actual visits are different explanations that need a fixed release vintage.

Start by reading the version

A changed tourism figure does not by itself mean that visits increased or declined. The process producing historical values may have been revised. Begin with a small exercise identifying the release and measure. Truthound can check keys and categories in the selected extract. It does not infer the cause of the change or evaluate a destination’s tourism performance from a difference between two files.

The release revises a historical interval

In the September 2026 data release published on 29 October 2026, New Zealand’s MBIE explained revisions to historical monthly unique visitors in Tourism Volumes and Flows. January 2024–March 2026 values were aligned with the upgraded mobile-data source used from April 2026. A new publication and a new observation therefore have different meanings that should remain visible in a receiving analysis.

Compare once within a single vintage

A proposed introductory extract retains release vintage, geographic level, observation month, visitor measure and source generation. Start with one measure for one region and a year-on-year comparison inside that vintage. Do not mix an old download’s past value with the newest file’s current value. Naming the chosen vintage makes the comparison recoverable when the publication changes again later.

Unique still has a geographic scope

The same person visiting several regions can count once in each region. Adding regional unique visitors does not produce a national unique-person count. Visitor days and unique visitors are also different measures. Keep the definition beside the number for someone learning to use the series, rather than expecting a familiar word such as unique to explain the counting boundary by itself.

Check the selected key and metric

Truthound Core Python SDK 3.3.4 (open-source SDK, Polars backend) documents MultiColumnUniqueValidator for an approved region-month-measure key inside the selected vintage. ColumnPairInSetValidator could check operator-approved measure and geography-level pairs. Staff define keys, categories and missingness first. Passing those rules does not establish the accuracy of the mobility model or explain the cause of real travel behaviour; it preserves the contract for the comparison being attempted.

Retain an existing duplicate mask when appropriate

pandas 3.1.1 (general-purpose DataFrame library with project-owned checks) DataFrame.duplicated can retain checks using the same approved key. Confirm that legitimate coexisting vintages are not being deleted as accidental duplicates. Compare Truthound through who can explain and correct an exception. Replacing an interface does not make regional unique counts mutually distinct at national level, and documentation supplies no evidence that either library improves the underlying tourism estimates.

Choose a useful first understanding

The first result should be an explanation of which vintage and measure were compared, rather than merely another visitor-growth percentage. Hold interpretation if either definition is missing. Extend to other regions or periods once that explanation works. Distinguishing a change in data production from a change in travel provides a stronger starting point for policy or business discussion than treating every revised value as new visitor behaviour.

Conditions

Separate release vintage, source generation, measure and geography in the tourism extract.

Alternatives

pandas 3.1.1 duplicated can retain a composite-key check within one release vintage.

Limitations

  • Mixed-vintage change is not evidence of more visits; regional uniques are not national unique people.

Evidence

Event checked
Subject checked
Modified