Check what each figure measures
Two figures about the same industry do not necessarily describe the same reality. One source may measure production, another retail sales, and a third the revenue of companies operating across several activities.
Record four elements for each value: definition, period, unit and geography. Add the data status where stated: observed, estimated, provisional or revised.
Distinguish the publication date from the observation period
A recent publication may contain a series ending several years earlier. The date on the cover does not turn those observations into current-year data.
On a company page, an investment announced on a given date must not be treated as money already spent without further information. A revenue target must not be presented as an actual result.
Check units and conversions
A conversion changes how data is expressed, not its scope. One hectolitre equals one hundred litres. Converting millions of hectolitres into millions of litres therefore multiplies the value by one hundred.
French production of syrups for consumption in our series analysis reached 2.031832 million hectolitres in 2023, a provisional figure, equivalent to 203.1832 million litres. The source is the table published by Agreste in Graph’Agri 2025, page 157.
Identify a break in series
A change in methodology, coverage or category can make two periods incomparable. In the syrup example, the series covers sales before 2012 and production from 2012 onwards. The article's chart starts in 2012 to maintain a consistent scope.
A change calculated across this break would mix two different measures. The solution is to separate the periods and explain the difference, even if this limits the length of the series presented.
Address differences between sources
When two publications give different values, do not simply choose the most recent one or the one supporting your argument. Compare their definitions and coverage. A difference may reflect an excluded activity, a different geography or a revision.
If the values cannot be reconciled, present them separately with their context. Missing data remains missing data; it does not become zero.