How did later sections change letters after the split?

Later sections changed letters after the split in ISIC Rev. 5, UN Statistics Division, because inserting a new section in the middle of the alphabet pushed the sections that followed it to later letters. The total reached 22 sections, A to V, and a section letter that meant one block in Rev. 4 may therefore mean a different block in Rev. 5. Any section level comparison across revisions must remap letters through the correspondence material, not carry letters across unchanged.

Why did one extra section move so many letters?

Alphabetical order is a convenience, not a meaning, and that is exactly why it can mislead. When Rev. 5 inserted a new section in the middle of the alphabet, every section after the insertion point shifted one letter later. The content of those later sections did not change merely because their letters did, but their labels did, and a label is what most readers see first. The total of 22 sections, A to V, is the result of that insertion plus the sections that already existed. A reader who opens a Rev. 4 table and a Rev. 5 table side by side therefore faces a trap: the same letter can stand for different blocks in the two revisions, and different letters can stand for blocks that are, in substance, continuous. Only the correspondence material resolves that puzzle, because it maps blocks by their definitions, not by their letters.

What goes wrong when a letter is carried across revisions unchanged?

The most common error is also the quietest. A writer copies a section letter from an old report into a new paragraph, or pastes an old chart beside a new one, and nothing looks wrong, because letters carry no warning flag. The comparison then proceeds as if the two sides measured the same block, when in fact the later letter may cover a different set of activities, or the same activities may now sit under a later letter. Errors of this kind survive proofreading, because the sentence is grammatically sound and the letter is correctly typed. They are caught only by a deliberate remapping step: for each letter in the old table, find the corresponding block in the Rev. 5 structure, read its definition, and record the new letter beside the old one. Skipping that step does not save time. It converts a small mapping job into a published comparison that cannot be defended when a checker asks what each side actually counted.

How can a long time series be read when its section letters shift?

Consider an analyst asked to show a decade of section level figures that straddle the revision. The analyst cannot simply extend the old series with new letters, nor simply rename the old letters to match the new ones. The workable approach is slower and more transparent. First, the analyst writes each section block in words, using the definitions from its own revision. Then the analyst uses the correspondence material to align the blocks, noting where a block maps cleanly, where it splits, and where it merges with a neighbour. Where alignment is incomplete, the analyst aggregates to a level where the blocks do line up, and marks the point of change in the series with a visible break note. The published table then shows its own history: readers can see where the letters changed, why they changed, and what was done about the change, instead of having to guess whether a jump in the line is real movement or a relabelling effect.

What check keeps a later section letter honest in Rev. 5?

Before quoting any section level figure from the later part of the alphabet, check four items together. Name the version, ISIC Rev. 5, UN Statistics Division, and do not let a Rev. 4 letter pass silently as a Rev. 5 one. Name the level, because a section total and a division figure inside it answer different questions. Name the reference period, because adoption dates can fall inside a series and create a break. Name the source of the classification, so a later reader can open the same UNSD structure and follow the same mapping. A survey team applies the same discipline earlier in the chain: questionnaire, coding manual, and published table must all point to the same definitions, or results drift between collection and publication. Where detail cannot be justified from the evidence, reporting at the broader level that is clearly correct is the stronger choice, not the weaker one.

For this topic above all, remember that this guide deals with statistical classification only: it does not fix any reader’s code, principal activity, or registration for them, and a code belonging to a particular record must be confirmed with the national statistical institute or business register that holds the record.