Where Is the Boundary Between Agriculture and Manufacturing in Rev. 5?

In ISIC Rev. 5, published by the UN Statistics Division, Section A covers primary production, while Section C covers manufacturing that transforms inputs into new products. Growing wheat is agriculture, while milling it into flour in a separate industrial process is manufacturing. Pressing olives on a farm for its own oil can raise a boundary question, and statistics resolve such cases by the nature and scale of the transformation and by principal activity.

What changes when a product is transformed?

Section A covers primary production, while Section C covers manufacturing that transforms inputs into new products. The difference is not the material. It is what happens to it. Growing wheat leaves wheat as wheat, in its natural or near natural state, apart from basic handling such as cleaning, grading, or drying. Milling that wheat into flour creates something new: a product with different properties, uses, and shelf life, made through an industrial process. The wheat went in. Flour came out. That change of character is the boundary, in its simplest form.

Basic handling deserves attention, because it is where readers most often hesitate. Cleaning grain, grading fruit by size, or drying a crop for storage does not transform it. The product remains recognisably itself, prepared for sale or storage rather than remade. Manufacturing, by contrast, changes character: milling, canning, fermenting at industrial scale, or assembling are all ways of making a new product from an input, and they belong on the manufacturing side even when they happen close to the farm gate.

Wheat, olives, and milk: three boundary tests

Growing wheat is agriculture, while milling it into flour in a separate industrial process is manufacturing. The two steps can happen on neighbouring premises, or even under one ownership, and still belong apart statistically, because the work differs. One step grows. The other transforms.

Pressing olives on a farm for its own oil can raise a boundary question, and statistics resolve such cases by the nature and scale of the transformation and by principal activity. Small scale pressing tied closely to the farm’s own olive output sits near the boundary, and the UNSD notes must be read carefully rather than guessed at. Scale matters here: a press that serves the farm’s own harvest is a different proposition from a press that buys olives from many growers and runs as an oil factory. Nature matters too: the degree to which pressing transforms the olive, and the industrial character of the process, both weigh in the reading.

Milk gives the cleanest contrast of all. A dairy farm selling raw milk illustrates agriculture: the farm’s work is keeping the herd and drawing milk from it. A factory pasteurising, bottling, and branding milk from many farms illustrates manufacturing: its work is treating, packaging, and presenting milk collected from elsewhere, through industrial routines. The milk is the same starting material, while the activity differs, and that difference, not the material, decides the side of the boundary.

Why does a business sometimes change sides?

A business that changes its main work over time may need a new heading, even though its name, premises, and customers have stayed the same. A farm that begins milling flour for sale at scale, or a dairy that builds a pasteurising plant taking milk from neighbours, has added a manufacturing enterprise to its farming. Its headline may shift as the balance of its work shifts, and that shift is not a punishment or a redefinition of the business. It is statistics following the work as it changes.

How should a borderline case be read?

On farm processing that is small and tied to the farm’s own output needs careful reading of the UNSD notes, not a guess. The notes carry inclusions and exclusions that a heading title cannot, and borderline cases usually turn on a detail buried in those notes: whether the process is listed as part of primary production, whether scale or industrial character is mentioned, or whether the output is sold as a farm product or as a manufactured good. Describing the work in plain words first, then looking for the heading whose description matches that work most closely, is slower than picking a familiar title, and far more reliable.

What does the wider structure add?

For boundary between agriculture and manufacturing, Rev. 5 published by the UN Statistics Division contains 22 sections from A to V, 87 divisions, 258 groups, and 463 classes in total, and it was endorsed by the UN Statistical Commission in March 2023. That detail is why a guide like this can only describe the boundary in general terms. The exact placement of a particular process lives in the class descriptions, where inclusions and exclusions are spelled out, and readers who need certainty must go there, and then to national guidance for local application.

This guide describes the boundary in general terms. It does not decide any individual farm or factory case, which needs the UNSD structure and national guidance. It is statistical information only, and it does not determine a reader’s code, principal activity, or registration; a formal decision for a real business rests with the national statistical institute or business register that holds the record.