What is a class in ISIC Rev. 5?

A class in ISIC Rev. 5, UN Statistics Division, is the finest level of the international classification, numbered with four digits, and Rev. 5 contains 463 classes. Classes are where the detailed definitions, includes lists and excludes lists do most of their work, because the differences between neighbouring classes can be narrow. Most code-level work in datasets, correspondence tables and methods notes happens at class level for that reason.

Why is the finest level also the most demanding to use?

A class is the finest level of the international classification, numbered with four digits, and Rev. 5 contains 463 classes. Fineness brings precision, but it also brings narrow neighbours. The differences between neighbouring classes can be slight, turning on a phrase in an explanatory note, an item in an includes list, or a redirect in an excludes list. A reader who works from titles alone will therefore make more mistakes at class level than at any broader level, because titles at this level often share words while their definitions draw a sharp line between them. The class level rewards slow reading, and it punishes skimming, in proportion to its detail.

That demand is also why most code level work in datasets, correspondence tables and methods notes happens at class level. Datasets need a common finest grain so that different sources can be combined and compared. Correspondence tables need to map one classification to another at the level where the mapping is actually defined, and that level is usually the class. Methods notes need to name the class so that a later reader can reproduce the coding decision. None of that work is possible at division or group level alone, because broader levels hide the very distinctions the work is about. The finest level is demanding precisely because it is where the classification does its real work.

What work happens at class level in datasets and correspondence tables?

In a dataset, class codes let an analyst aggregate upward with confidence. Class figures sum to group totals, group figures sum to division totals, and division figures sum to section totals, because every class belongs to exactly one group, every group to exactly one division, and every division to exactly one section. That chain fails if a class code is wrong, because the error propagates upward into every total that contains it. Coding at class level therefore carries a responsibility that coding at a broader level does not: a single misplaced class can distort several published totals at once, while a broader figure, though less detailed, at least confines its error to its own level.

Correspondence tables add a second kind of class work. When a national classification, a product classification, or an earlier revision is mapped to Rev. 5, the mapping is usually specified class by class, because broader mappings would hide splits, merges and moves that change the meaning of the figures. A reader who uses a correspondence table should therefore check it at class level, even when the final table will be published at division level. The published total may be broad, but its construction is detailed, and a mapping error at class level will survive aggregation and appear, disguised, in the broad total.

How does a coder choose between two neighbouring classes?

Describe the activity in words first, without borrowing the title of either candidate class. Then read the explanatory notes for both classes, side by side, looking for the phrase that distinguishes them: a difference in the kind of work, in the method of production, or in the scope that one note includes and the other excludes. Then check the excludes lists, because a redirect in one class often points directly to the other, and that redirect is frequently the deciding evidence. If neither note fits clearly, the honest step is to stay at the broader group level that is clearly correct, and to record the question, rather than to pick a fine category that cannot be justified from the evidence available.

That habit also protects against a common shortcut: choosing the class whose title sounds most familiar. Everyday words are given technical definitions in ISIC Rev. 5, UN Statistics Division, and the familiar word in a title may mean something narrower, or broader, than it means in everyday speech. A coder who trusts familiarity will sometimes land in the right class by luck, and sometimes in a neighbouring class by the same luck, with no way to tell which outcome occurred. A coder who trusts the notes will land in the right class by reasoning, and will leave a trail a colleague can follow and, if necessary, correct.

This guide explains the finest level of a statistical classification, and it does not assign that level to any individual reader. For a class, read this guide as statistical classification only: it assigns no code, principal activity, or registration to any reader, and a specific record is confirmed by the national statistical institute or business register holding it.