Points, polygons and the four-hectare rule

Almost every due diligence statement that comes back for correction comes back because of geometry. Not because the supplier was hiding anything, and not because the land was cleared — because a coordinate had four decimal places instead of six, or because a 6.5-hectare block arrived as a single point, or because a polygon's last vertex missed its first by a metre and a validator refused to close the ring.

None of that is a judgement about deforestation. It is a data problem that gets treated as a compliance failure, and it is worth understanding precisely, because the rules are less complicated than the folklore around them suggests.

What the regulation actually asks for

Two provisions do the work. The obligation lives in Article 9(1)(d), which requires operators to collect the geolocation of all plots of land where the relevant commodity was produced, together with the date or time range of production. The definition that decides what "geolocation" means sits somewhere else entirely, in Article 2(28) — and that split is why so many teams read the obligation without ever reading the rule that governs it. Both are in the consolidated text on EUR-Lex.

Article 2(28) describes geolocation as the geographical location of a plot of land, given as latitude and longitude coordinates using at least six decimal digits. For plots above four hectares used to produce a relevant commodity other than cattle, it requires polygons with enough points to describe the perimeter of each plot.

The unit being described is defined separately again, at Article 2(27): a plot of land is land within a single real-estate property as recognised by the law of the country of production, with conditions homogeneous enough that the deforestation risk across it can be assessed as one thing. That last clause is the operative part, and it is regularly missed. A single legal parcel that contains both a mature coffee block and a recently converted section is not one plot for these purposes, because you cannot assess it as one.

The four-hectare line, and what sits on either side

Four hectares is 40,000 square metres — a square roughly 200 metres on each side. It is a small farm by industrial standards and a large one by smallholder standards, which is exactly why the threshold lands where it does: it separates holdings that can reasonably be described by one coordinate from holdings where one coordinate tells you almost nothing.

The threshold applies per plot, not per supplier and not per consignment. A cooperative delivering from four hundred farms is not making one four-hectare judgement; it is making four hundred of them. In practice that means the collection process has to capture area alongside position, because you cannot know which obligation applies until you know how big the parcel is — and a supplier's own estimate of their acreage is not a reliable input.

Six decimal places is not a formatting preference

The precision requirement gets treated as pedantry until someone works out what it means on the ground. A degree of latitude is about 111 kilometres, so each decimal place divides that by ten.

Eleven metres sounds tolerable until you consider what it is being used for. The coordinate is not decoration; it is the input to a deforestation check against the 31 December 2020 baseline. An eleven-metre error at a parcel boundary can put your point inside a neighbouring block that was cleared in 2022, and you have then made a positive, documented assertion about land you never sourced from. That is materially worse than supplying nothing, because it is now in the record with your name against it.

Truncation is usually the culprit rather than bad fieldwork. Coordinates get collected at full precision on a phone, exported to a spreadsheet, and silently rounded by a column format on the way through. It is worth checking the file that actually leaves your organisation rather than the app that captured it.

What a polygon has to be, geometrically

The regulation asks for enough points to describe the perimeter and stops there. The practical constraints come from the EU Information System, which accepts geometry as GeoJSON in WGS-84 (EPSG:4326) and validates it before a statement is accepted. Those are two different sources of obligation, and it is worth keeping them apart in your own documentation: the first is law, the second is a system that will reject you on a Tuesday afternoon.

In practice a polygon needs to be a closed ring whose first and last coordinate are identical, without self-intersections, and without interior holes. Each of those has a characteristic failure mode:

• A ring that is nearly closed — the last vertex a metre from the first — renders perfectly on a map and fails a topology check. This is the single most common rejection we see.
• A self-intersecting outline, usually produced by walking a boundary and crossing your own track, has no well-defined interior, so its area cannot be computed and it cannot be tested against anything.
• Interior holes, sometimes used to exclude a house or a watercourse, are not part of the expected shape and are better handled by splitting the parcel.
• Coordinate order reversed into longitude-latitude puts a West African cocoa farm somewhere in the Gulf of Guinea. It validates cleanly, which is what makes it dangerous.

Cattle are different, in both directions

Cattle sit outside the polygon rule entirely. What Article 9(1)(d) asks for is the geolocation of the establishments where the animals were kept, and those are supplied as points regardless of how much land the establishment covers.

That is easier in one respect and considerably harder in another. There is no perimeter to survey — but an animal that moved between three holdings before slaughter generates three establishments to identify, and the traceability question becomes one of records rather than geometry. Beef supply chains that have solved their mapping problem often have not solved this one at all.

A low-risk origin does not remove the geometry

This is the most expensive misunderstanding in the area, and it has become more common since the country benchmarking list was published. Sourcing from a country classified as low risk gives access to simplified due diligence under Article 13, which means the risk assessment and mitigation steps in Articles 10 and 11 are not required in the absence of a substantiated concern. It does not touch the information-collection duty in Article 9.

Put plainly: a low-risk origin changes how much analysis you do with the plot data. It does not change whether you have to collect it, and it does not change what shape it has to be. Teams that read the benchmarking announcement as an exemption and stood their data collection down have the furthest to travel before December.

What to do before the deadline

1. 1

   Measure area before you decide on geometry

   The point-or-polygon question cannot be answered from a supplier's stated acreage. Derive area from the geometry itself, then apply the threshold.

2. 2

   Validate at collection, not at submission

   A ring that will not close is trivially fixable while someone is standing in the field and close to impossible to fix in November from another continent.

3. 3

   Check precision on the file that leaves your organisation

   Truncation happens in the export, not the capture. Inspect the artefact you actually send.

4. 4

   Separate structural validity from correctness

   Run the topology checks, then check the shape against imagery. The two answer different questions and you need both.

None of this is conceptually difficult. It is a data quality exercise carried out at a scale most supply chains have never had to attempt, against a deadline, with suppliers who have no particular reason to care about the difference between five and six decimal places. Starting from the geometry — rather than from the paperwork — is what makes it tractable.

Published June 16, 2026 · Updated August 12, 2026 · Last reviewed August 12, 2026 against the sources listed above.