NewsBeagle Blog · Practical guides
Provisional vs Revised Statistics: Compare the Right Versions
Learn to distinguish a revised estimate from a new period's result, with a version ledger and worked example that avoids a misleading growth claim.
A changed estimate is not automatically a changed real-world outcome
A provisional statistic is an early estimate that its producer may update. A revised figure can describe the same place and period using more complete information or different methods. A new release date does not necessarily mean the figure measures a new month. Before interpreting two numbers as growth or decline, identify the reference period each number describes.
The Office for National Statistics' population and migration revisions policy explains that updates can incorporate additional data, replace assumptions or improve methods; it distinguishes scheduled revisions from corrections of identified errors. That is a policy for those statistics, not a timetable or rule for every dataset. Read the particular producer's release notes to learn why your series changed.
Keep a version ledger beside the headline
For each number, record six fields: measure and unit; geography; reference period; release date; version status; and source link. Add definitions that affect comparability, such as whether a count means arrivals, visits or unique people. If the page includes a series identifier, save it too. Two similar labels do not guarantee the same population or method.
Next, make two different comparisons. A revision comparison holds the reference period constant and changes the release version. A period comparison holds the definition and version basis as consistent as possible and changes the reference period. Write down which comparison you intend before calculating a percentage. This simple separation prevents an update to last month's estimate being mistaken for this month's performance.
Worked example: the misleading 20 percent increase
Suppose a fictional bulletin first estimates 1,000 visits during June. A later bulletin revises June to 1,200 and publishes a first July estimate of 1,100. Comparing the two June versions gives a revision of 200 visits, or 20 percent of the first estimate. It does not show that visits grew by 20 percent from June to July: both June figures describe the same month.
Using the later bulletin's June estimate, July's 1,100 is 100 below June's 1,200, about 8.3 percent lower. That period comparison is still provisional if July can be revised. Comparing July with the superseded June figure would instead suggest a 10 percent increase. The arithmetic is valid in both calculations, but the choice of versions changes the story. These numbers are hypothetical, not tourism statistics for Thailand or any other country.
Check whether the whole series has changed
Do not replace one number in a chart and assume its neighbours remain comparable. Open the revised table and its notes: earlier periods may also have changed. If units, boundaries or methods differ, explain the break or use the producer's comparable back series where one exists. If you cannot establish a consistent basis, do not publish a precise growth claim from mismatched figures.
Save the release URL and access date with your working notes. Where the publisher provides previous versions, keep links to both the earlier and current releases. If it overwrites a table without an accessible history, state that you cannot reconstruct the earlier version rather than treating an undated screenshot as an official record.
Our editorial approach
This guide is written by the NewsBeagle Editorial Desk to help readers compare attributed coverage and find original reporting. See our editorial policy and publisher directory.